<<

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ Atmospheric © Author(s) 2009. This work is distributed under the Creative Commons Attribution 3.0 License. and Physics

A case study of production, , and the budget in City

E. C. Wood1, S. C. Herndon1, T. B. Onasch1, J. H. Kroll1, M. R. Canagaratna1, C. E. Kolb1, D. R. Worsnop1, J. A. Neuman2, R. Seila3, M. Zavala4, and W. B. Knighton5 1Aerodyne Research, Inc., Billerica, Massachusetts, USA 2Earth System Research Laboratory, NOAA, Boulder, Colorado, USA 3National Exposure Research Laboratory, Environmental Protection Agency, Research Triangle Park, North Carolina, USA 4Molina Center for Energy and the Environment, La Jolla, , USA 5Department of Chemistry and , Montana State University, Bozeman, Montana, USA Received: 14 July 2008 – Published in Atmos. Chem. Phys. Discuss.: 19 August 2008 Revised: 25 March 2009 – Accepted: 31 March 2009 – Published: 7 April 2009

Abstract. Observations at a mountain-top site within the scale of one day. A new metric for ozone production effi- basin are used to characterize ozone production ciency that relates the dilution-adjusted ozone mixing ratio and destruction, nitrogen speciation and chemistry, and to cumulative OH exposure is proposed. the radical budget, with an emphasis on a stagnant air observed on one afternoon. The observations compare well with the results of recent photochemical models. An ozone 1 Introduction production rate of ∼50 ppbv/h was observed in a stagnant during the afternoon of 12 March 2006, which is 1.1 Megacities and air among the highest observed anywhere in the world. Approx- imately half of the ozone destruction was due to the oxida- Urban areas are often characterized by the presence of large tion of NO2. During this time ozone production was emissions of , including volatile organic com- VOC-limited, deduced by a comparison of the radical pro- pounds (VOCs), nitrogen oxides (NOx≡NO+NO2), duction rates and the formation rate of NOx oxidation prod- dioxide (SO2), and particulate matter (PM). Many of these ucts (NOz). For [NOx]/[NOy] values between 0.2 and 0.8, compounds (e.g., and diesel PM) are directly toxic - HNO3 typically accounted for less than 10% of − to humans (Khan, 2007; McCreanor et al., 2007; Yauk et NOz and accumulation-mode particulate (NO3(PM1)) al., 2008). These primary pollutants react in the , accounted for 20%–70% of NOz, consistent with high ambi- forming secondary pollutants such as ozone, , ent NH3 . The fraction of NOz accounted for − nitric , and secondary PM. The spatial distribution of by the sum of HNO3(g) and NO3(PM1) decreased with pho- secondary pollutants and their direct effect on public health, tochemical processing. This decrease is apparent even when vegetation, and the ’s radiative balance can be quite dif- dry deposition of HNO3 is accounted for, and indicates that ferent than that of their precursor primary pollutants. Ozone HNO3 formation decreased relative to other NOx “sink” pro- (O3), in particular, plays a central role in tropospheric chem- cesses during the first 12 h of and/or a sig- istry, as it is both a product and initiator of photochemistry nificant fraction of the nitrate was associated with the coarse as well as a potent . Ozone is one of the US size mode. The ozone production efficiency of NOx EPA’s six criteria air pollutants subject to national ambient air on 11 and 12 March 2006 was approximately 7 on a time quality standards (NAAQS) created to protect public health. Numerous air basins across the world, spanning urban, sub- urban, and rural areas, have difficulty meeting relevant air Correspondence to: E. C. Wood quality standards for O3. Understanding the formation of O3 ([email protected]) and the transformations of its precursors (VOCs and NOx) is

Published by Copernicus Publications on behalf of the European Geosciences Union. 2500 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City crucial for reducing on both urban and regional founding influence of . Secondary organic aerosol scales. Over the past several decades great effort has been formation at this site is analyzed and described in two related focused on characterizing air pollution and atmospheric pho- manuscripts (Wood et al., 2009; Herndon et al., 2008). The tochemistry in the US and Europe (Carslaw et al., 2001; Em- observations and inferences regarding tropospheric chem- merson et al., 2007; Kleinman, 2005; Nunnermacker et al., istry are compared to the predictions of several photochem- 1998; Ryerson et al., 2001; Tressol et al., 2008). However, ical models (Lei et al., 2007; Madronich, 2006; Tie et al., there are few studies of tropospheric photochemistry in de- 2007). Such observational-based characterizations of pho- veloping megacities – where much of the projected increases tochemistry, even if focused on a short period of time, are in population and emissions over the next several crucial for testing our understanding of the underlying pho- decades are expected to occur (Molina et al., 2004). tochemical processes that control secondary air pollution. During the 2006 MILAGRO (Megacity Initiative: Local And Global Research Observations) campaign, Mexico City 1.2 Urban photochemistry overview was used as a test case for developing megacities. Mexico City is currently the third most populous urban agglomer- is produced by the oxidation of VOCs ation in the world, with a population of 19 million (UN, (volatile organic compounds) in the presence of nitrogen ox- ≡ 2008). Although the air quality in Mexico City has improved ides (NOx NO+NO2) and . Central to understand- greatly over the past 10 years (de Foy et al., 2008), it remains ing ozone production is the photostationary state formed be- highly polluted, with ozone concentrations frequently ex- tween NO, NO2, and O3 in sunlight: ceeding 100 ppbv. The motor vehicle fleet, consisting of over NO + O3 −→ NO2 + O2 (R1) 4 million cars and trucks, is older than the US fleet, and ap- 3 proximately 30% of light-duty vehicles do not have NO2 + hν(λ < 400 nm) −→ NO + O( P) (R2) functioning 3-way catalytic converters (CAM, 2008). The 3 + + −→ + latitude (19◦ 24◦ N) and of the Mexico City plateau O( P) O2 M O3 M (R3) (2.2 km above sea level) to elevated actinic flux levels Net reaction: null throughout the year. The high altitude also tends to cause NO can also be oxidized to NO2 by compounds other than fuel-rich in the vehicle fleet. O3, most importantly hydroperoxy radicals (HO2) and or- Inferences about tropospheric photochemistry based on ganic peroxy radicals (RO2, where “R” represents an organic measurements from stationary sites are often complicated ): by the fact that the observed changes in pollutant concentra- tions over time are a function of not just chemistry but also HO2 + NO −→ OH + NO2 (R4) and emissions patterns. Additionally, the air + −→ + masses observed usually contain a range of aged and fresh RO2 NO RO NO2 (R5a) pollutants, since primary emission sources of NO , VOCs, x During the day, most of the NO2 formed from R4 and R5a and PM continue to impact most stationary sites through- undergoes photolysis, leading to the creation of ozone (R2 out the day. Measurements on board instrumented aircraft and R3). When discussing ozone production quantitatively, are often better suited for tracking the evolution of an air it is useful to use the concept of Ox (“odd-”), which mass over time (Ryerson et al., 2001), though the first 3– in the is often defined as the sum of O3 and NO2 7 h of atmospheric oxidation following sunrise usually occur ([Ox]≡[O3]+[NO2]). NO2 is included because it acts as a in a shallow boundary layer near the urban surface where it reservoir of O3, formed by R1 or as a source of O3 (R4 and may be unfeasible to fly. chamber studies have pro- R5a). The sum of the rates of R4 and R5a is therefore equated vided insight into numerous atmospheric processes, but are to the instantaneous production rate of Ox: challenged by wall interactions and the difficulty in realisti- X cally mimicking atmospheric parameters such as actinic flux, P(Ox) = k5ai[RiO2][NO] + k4[HO2][NO] (1) low pollutant concentrations, and total atmospheric composi- i tion. Some atmospheric processes have not been adequately Although the right-hand side of Eq. (1) is often defined characterized by laboratory studies, e.g. secondary organic as P(O3) (i.e., the instantaneous production rate of O3 rather aerosol formation (de Gouw et al., 2005; Volkamer et al., than Ox), by defining it as P(Ox) instead there is no need 2006). This emphasizes the importance in quantifying such to account for reactions of NO2 subsequent to its formation, processes using atmospheric measurements. and the expression holds true at all light levels. We note that In this , we present measurements from Pico de Tres during the day, typically over 95% of the NO2 will undergo Padres, a unique mountain-top stationary site that is situated photolysis to form O3 (the remainder reacts to form HNO3 within the Mexico City basin but is minimally impacted by or other compounds), and thus the terms “Ox production” nearby emissions. The air masses observed on 12 March and “ozone production” are often nearly equivalent. How- 2006 were stagnant in the afternoon and provided an excel- ever, in the presence of large concentrations of NO, pho- lent opportunity to study ozone chemistry without the con- tochemically formed Ox may appear mainly in the form of

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2501

NO2, underscoring the advantage of considering Ox rather 1 2200 Elevation (km) than O3. A more inclusive definition of [Ox] includes O( D), 3.2 3 O( P), NO3, and N2O5, but these compounds are reasonably 2.8 T1 assumed to be minor contributors to Ox concentrations com- 2180 2.4 pared to O and NO during daytime. 2.0 3 2 PTP In fresh urban plumes with active photochemistry, the pro- 2160 T0 duction rate of Ox is often much greater than the total rate of the Ox losses, and thus ozone can accumulate in high concen- trations (>100 ppbv). Ox is fairly long-lived, with a lifetime 2140 PED range of about a day to weeks.

The sum of NOx and its oxidation products are known UTM Northing (km) STA 2120 as “” or NOy (NOy≡NO+NO2+HNO3+ − NO3(PM1)+organic +NO3+2N2O5+ HONO+. . . ), and the oxidation products alone are known as NOz (NOz≡ 2100 NOy−NOx). All NOz compounds, with the exception of HONO, NO3, and N2O5, are the result of three categories 440 460 480 500 520 540 560 of ROx−NOx reactions: UTM Easting (km) RO2 + NO + M −→ RONO2 + M (R5b) Fig. 1. Selected deployment locations for the mobile laboratory dur- NO2 + OH + M −→ HNO3 + M (R6) ing MILAGRO-2006. The figure is a topographical representation of the Mexico City Metropolitan Area. The sites “T0” and “T1” are RO2 + NO2 + M ←→ RO2NO2 + M (R7) the campaign supersites. This work focuses on measurements from Pico de Tres Padres (PTP). Santa Ana (STA) and Pedregal (PED) Aerosol nitrate is mainly formed as nitrate are the sites of two additional mobile laboratory deployments in the following reaction of HNO3 with NH3. In this paper, the south of the city. The urban core of the federal district is encom- products of R5b and R7 are collectively referred to as “or- passed by T0, PED, and STA. UTM=universal transverse mercator. ganic nitrates”, and comprise formed The border of the Distrito Federal is depicted by the grey dots. from R7 (most commonly peroxyacetyl nitrate, PAN) and multi-functional nitrates formed from R5b. Partition- ing of organic nitrates can form aerosol nitrate as well (Lim (http://www.asp.bnl.gov/MAX-Mex.html) that were hosted and Ziemann, 2005), though the extent to which this occurs by the Instituto Nacional de Petroleo and the Universidad in the atmosphere is currently not well quantified. Tecamac, respectively. PTP is ∼10 km northeast of T0 and ∼19 km southwest of T1. 2 Experimental section (NO) and total reactive nitrogen (NOy) were measured with a ThermoElectron model 42C chemilumines- Measurements were made aboard the Aerodyne mobile lab- cence sensor in conjunction with an external oratory, a panel truck outfitted with a suite of real-time in- (Mo) converter that was mounted to the roof of the mobile strumentation for gaseous and fine-particulate measurements lab. Measurements of NO were also made with an ECO (Herndon et al., 2005; Kolb et al., 2004). The mobile lab- Physics model CLD 88Y sensor. Both oratory was deployed during the MAX/MEX portion of the instruments were periodically calibrated with a standard tank MILAGRO-2006 campaign to various sites throughout the of NO (Scott Specialty ), diluted with NOx-free zero Mexico City Metropolitan Area (MCMA). Sites were cho- air using a ThermoElectron dynamic gas calibrator. The ef- sen based on the 2–5 day weather forecast and knowledge of ficiency of the Mo converter for NOy was quantified by cal- how synoptic scale weather patterns typically produce pre- ibrations with NO2 and n-propyl nitrate. Particulate nitrate vailing basin flow archetypes (de Foy et al., 2005; de Foy was assumed to be converted with the same efficiency as et al., 2008; Fast et al., 2007). The goal of these deploy- gas-phase species given the high surface area mesh design ments was to sample the urban outflow. Some of the mo- of the molybdenum converter (Williams et al., 1998). The bile laboratory deployment locations are depicted in Fig. 1. 1σ uncertainty of the NO 1-min data is the greater of 7% or Most of the data presented in this manuscript were recorded 0.5 ppbv (NO), and the uncertainty of the NOy data is esti- at Pico de Tres Padres (PTP), which is the highest point in mated as the greater of 15% or 1.5 ppbv. NOz concentrations the Sierra de Guadalupe range in the northern section of were calculated using the NOy, NO, and NO2 measurements the MCMA. The sampling altitude at PTP was 720 m above (i.e., [NOz]=[NOy]–[NO]–[NO2]). the MCMA valley elevation of 2.2 km. PTP is situated be- Two dual-laser Aerodyne tunable laser differ- tween “T0” and “T1”, two of the MAX/MEX supersites ential absorption spectrometers (TILDAS) using pulsed www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2502 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City quantum cascade lasers (Herndon et al., 2007; Nelson et al., Thirty-one were quantified using GC-FID from −1 2004) were used to measure NO2 (at 1606 cm ), CO (at whole air samples collected in stainless canisters (Seila −1 −1 −1 2100 cm ), HCHO (at 1722 cm ), HNO3 (at 1721 cm ), et al., 2001). The alkenes were ethene, , 1-, −1 and NH3 (at 965 cm ). The inlet for these two instruments trans-2-butene, cis-2-butene, 1-pentene, trans-2-pentene, cis- consisted of 1 m of heated 1/4” OD PFA tubing followed by a 2-pentene, 1-hexene, 1-heptene, t-3-hexene, 1-octene, t-4- heated teflon-coated cyclone (URG-2000-30ED; aerosol di- octene, 1-decene, isobutene, 3-methyl-1-butene, 2-methyl- ameter size cut-off of 0.8 µm at a flow rate of 8 liters per 1-butene, 2,4,4-trimethyl-1-pentene, 2-methyl-2-butene, 2,4- minute) and a siloxyl-coated orifice to reduce the sam- dimethyl-1-pentene, 2-methyl-2-pentene, 1,3-, cy- pling pressure. The estimated uncertainties for the NO2, CO, clopentene, cyclooctene, cyclohexene, , limonene, and HCHO measurements are 8%. In order to assess the in- camphene, beta pinene, and alpha pinene. let transmission for NH3(g) and HNO3, small flows of both Photolysis rate constants for O3, HCHO, , gases from permeation tube sources were introduced into the and were determined by a combination of measure- inlet four times. The ambient aerosol loadings (measured ment and model results. The photolysis rates calculated from with the aerosol mass spectrometer) during these standard the Scanning Actinic Flux Spectroradiometer operated by B. 3 additions were in the range 10 to 30 µg/m . The ambient Lefer and J. Flynn at the T1 site were used in conjunction ◦ ◦ temperature ranged from 16 to 22 C and the relative humid- with the NCAR tropospheric and visible ity ranged from 28% to 45%. For comparison, the range of model (TUV). Equation (2) was used to estimate photolysis conditions observed between 06:00 and 20:00 LT at PTP was jPTP(T UV ) rates:jPTP= ×jT1(measured) (2)where jPTP(TUV) 10◦ to 27◦C and 12% to 75% relative . Inlet losses jT1(T UV ) and jT1(TUV) are the model results and j(measured) is from of NH3 and HNO3 were on order 28% and 45% respectively; the T1 measurement. This approach uses TUV to account for the measurements have been proportionally corrected. Given the difference in elevation between PTP and T1 (850 m), for the size of the inlet losses and the limited range of ambient which aerosol and absorbing gases affect the optical trans- conditions under which standard additions were performed, mission. The hourly vertical profiles of ozone used in the we estimate an uncertainty of 50% for the HNO3 measure- model were generated from the measurements at T0 and PTP. ments (the NH3 measurements are not used quantitatively in Although no information regarding the NO2 and O3 column this analysis). Only NH3 and HNO3 measurements from PTP above the convective mixing layer was included, the system- are included in the analysis. atic errors associated with this are expected to cancel in the Ozone (O3) was measured with a dual-beam absorbance first factor of the right-hand side of Eq. (2). The total ozone photometer at 254 nm (2B Tech model 205), with an ac- column is characterized by the T1 measurements. The ratio curacy of 2%. Acetaldehyde, acetone, , ben- of jPTP(TUV)/jT1(TUV) depends on the specific molecular zene, and C3-benzene compounds (sum of C9H12 isomers photolysis rate, but for all of the compounds discussed in this and C8H8O isomers) were measured with a -transfer work, the altitude correction is less than 16%. mass spectrometer (Rogers et al., 2006). With the ex- The largest source of uncertainty in the photolysis rate ception of acetonitrile, concentrations were derived from constants by this approach lies in the fact that PTP is ∼19 km response factors determined from a calibrated gas stan- away from T1 and the cloud cover was not spatially identi- dard with an accuracy of 25%. Acetonitrile concentrations cal. A comparison of the measured radiometer data at T1 were determined from transmission corrected intensi- and an Eppley UV photometer used aboard the mobile lab- ties and equations derived from standard reactions kinetics oratory suggests qualitative agreement when cloud patches (Lindinger et al., 1998) assuming a constant of moved above the region. For the periods in which the pho- 4.74×10−9 ml −1 s−1 (Zhao and Zhang, 2004). tolysis rates were analyzed, there was minimal cloud cover Particulate nitrate (vacuum aerodynamic diameters be- (i.e., it was not cloudy). tween 60 and 800 nm–50% cut points) was measured with an Aerodyne compact time-of-flight aerosol mass spectrom- A second source of uncertainty in this approach stems eter (C-ToF-AMS) with an accuracy of 20% (Canagaratna from the differences in albedo and upwelling/downwelling et al., 2007; Drewnick et al., 2005; Liu et al., 2007). A radiation associated with the elevation and topographical dif- collection efficiency due to particle bounce of 0.5 was used ferences between T1 and PTP. The T1 measurement data is for all species during the MILAGRO study based on com- based on the downwelling radiation, however at PTP the pho- parisons with other aerosol instrumentation, including a co- tolysis rate in the sampled air had a contribution from re- located Scanning Mobility Particle Sizer (SMPS; TSI model flected light. We estimate that this contributes an uncertainty 3080) and recent laboratory studies (Canagaratna et al., of 4% in the calculated photolysis rates, based on the average 2007;Matthew et al., 2008). Size distribution measurements Mexico City albedo. from both the AMS and SMPS from this study and previous Additional sources of error in this approach are due to the measurements in Mexico City (Salcedo et al., 2006) indicate absence of modeled absorbers in the 850 m elevation dis- that the particulate measurements discussed here represent tance between T1 and PTP. Hourly profiles of ozone and PM1 mass loadings. NO2 were used in the model, though SO2, HONO, HCHO

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2503 and acetaldehyde were ignored. The bias introduced by ne- 200 glecting these absorbers would result in an underestimate of the photolysis rates at PTP since such absorption leads to 150 higher photolysis frequencies at higher . This un- (ppb) derestimate is likely most pronounced in the early morning 3 100 Ox (O3 + NO2) when the convective mixing layer height is below the altitude O O3 of PTP. 50 The combined uncertainties in the photolysis rates (ac- counting for the altitude, cloud coverage, albedo, and un- 0 CO NO known absorbers) is estimated as 25%. 80 y 1500 NO2 All stated accuracies reflect 1σ uncertainty. All reaction NO 60 z rate constants used are those from the latest JPL recommen- NO

(ppb) 1000 y

dations (Sander et al., 2006). One-minute averages are used 40 CO (ppb) for all measurements presented, with the exception of the NO 500 VOC canister measurements which were 30 min averages. 20 The NO2 data are one minute averages recorded every three 0 0 minutes, thereby affecting the Ox and NOz values too. All 08:00 10:00 12:00 14:00 16:00 18:00 linear correlation fits were calculated using a least-squares algorithm that accounts for the uncertainty in both variables. Local time (CST)

Fig. 2. Time series of several species at PTP on March 12, 2006. The time period between 12:15 and 13:15 was characterized by 3 Results and discussion stagnant air and is used to infer the production rates of Ox and NOz. The gap in the data between 08:30 and 10:00 is due to instrument 3.1 Air masses observed at PTP maintenance.

Because the mobile laboratory payload did not include any day accumulation of pollutants is rare in Mexico City due to direct measurements of the vertical structure of the atmo- efficient venting (de Foy et al., 2006). sphere, the measurements of the convective mixing layer PTP was above the nocturnal boundary layer during most heights at T0 and T1 (Shaw et al., 2007) have been exam- of each night. With the exception of a few episodes in which ined as a proxy. Typically, during clear-sky insolation, the fresh emissions were brought to the site by strong winds ∼ mixed layer grew to a height of 1 km above T0 by 11:30 or when trapped residual layers advected past the site, the local time (LT). CO mixing ratios increased sharply at PTP nighttime measurements were indicative of a clean resid- near 09:30. That pollution reached PTP before the mixing ual boundary layer. For example, the mean nighttime mix- layer height reached the elevation of PTP (as measured at T0 ing ratios of CO, O3, NOy and CH3CN (acetonitrile) be- and T1) is most likely due to differential heating and upslope tween 02:00 and 06:00 on 12 March were 132, 44, 0.6, and winds. 0.2 ppbv, respectively. The organic aerosol loading at night The altitude and horizontal location within the city and the was 1.1 µg m−3 (STP, 273 K and 1 atm) and was highly oxi- lack of short “spikes” in the time series of all species indi- dized. cate that PTP was not greatly affected by nearby emissions, Much of this work focuses on data from 12 March 2006, and thus was a near-ideal stationary measurement location in which there was a one-hour pause in the rise of the mix- for sampling mixed urban emissions during their first 2 to ing layer in the afternoon. Figure 2 shows a time series of 12 h of photochemistry. This is in contrast to most ground CO, O3,Ox, NOy, NO, NO2, and NOz concentrations ob- sites within cities, which are usually close (within 100 m) to served at PTP on Sunday, 12 March 2006. Concentrations of traffic and other pollution sources. Based on the measured all species rose sharply starting at ∼09:00 CST, as upslope wind speeds, the emissions from the closest neighborhood winds transported pollutants from below the sampling eleva- (∼4 km to the south) would have had on the order of 15– tion to PTP. Concentrations of CO (which is relatively pho- 60 min to mix into the urban plume. On 12 March 2006 (the tochemically inactive) reached an initial peak at 10:40 due to day highlighted in this work), the transport time from other upslope winds, though the mixing layer depth at T0 did not parts of the city was on the order of 1–5 h, whereas vertical reach the elevation of PTP (720 m above the city elevation) mixing time scales in the afternoon are typically less than 30 until ∼11:30 as described earlier. [CO] decreased between minutes (Stull, 1988). Therefore, as the mixing layer grew 11:00 and 12:00 as the mixing layer rose above PTP and the above PTP, the sampled air masses were typically an accu- effect of dilution overweighed the flux of CO from below. mulation of pollutants from many sectors of the city mixed both vertically and horizontally over several hours. Multi- www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2504 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City

The measurements of the convective mixing depth above Table 1. Quantification of the destruction (loss) rate of Ox be- T0 show no increase between 12:00 and 13:00 on 12 March tween 12:15 and 13:15 on 12 March 2006. Mixing ratios used for 2006 (Shaw et al., 2007). The urban emission inventory these calculations: [O3]=160 ppb (measured), [H2O]=7 ppth (mea- 6 3 (CAM, 2008), which apportions on-road emissions by hour sured), [HO2]=40 ppt (estimated), [OH]=6×10 molec/cm (esti- of day, indicates that the amount of CO emitted between mated). A boundary layer height of 0.9 km and a deposition ve- 12:00 and 13:00 is equal to 15% of the amount emitted be- locity of 0.4 cm/s were used to calculate the dry deposition rate of tween 05:00 and 12:00. This should to an increase O3. The oxidation rate of NO2 was calculated as 0.8×P(NOz). of [CO] by approximately 15% above the background mix- ing ratio if there is no mixing with the residual layer aloft Reaction Rate (ppbv/h) caused by a rising boundary layer. The observed enhance- O(1D)+H O 1.0 ment in [CO] between 12:15 and 13:15 was 14% (calcu- 2 HO2+O3 0.8 lated as ([CO] : –130 ppbv)/([CO] : –130 ppbv) to ac- 13 15 12 15 OH+O3 0.2 count for the background CO mixing ratio of 130 ppbv). The NO2→NOz 6.0 favorable comparison between the observed and predicted in- O3+VOCs 0.7 crease in [CO] complements the T0 measurements showing O3 dry deposition 2.6 a pause in the rise of the mixing layer between 12:15 and NO3+hν→NO+O2 0.3 13:15. This analysis is not sensitive to inaccuracies in the Total 11.6 absolute magnitude of CO emissions; it is sensitive only to the reported timing of CO emissions between the early morn- ing and afternoon, since the quantity used is the relative in- crease in [CO] rather than the absolute mixing ratios. This merous aspects of tropospheric chemistry. The average wind time period was unlike most other days at PTP, when bound- speed during this time period was 1.5 m/s from the south- ary layer dynamics and advection played much greater roles east, so the air sampled spanned less than a 6 km horizontal in determining changes in pollutant concentrations. range – most of which is unpopulated land along the moun- A comparison of the ratios of several primary pollutants tain slope, with no urban emission sources. For comparison, at PTP and T0 (a site greatly impacted by fresh urban emis- the distance between T0 and PTP is ∼10 km. Other days sions) indicates that the air observed at PTP was represen- were excluded from this analysis because of at least one of tative of mainly urban emissions and was not greatly af- the following two filtering criteria: 1) the wind speeds were fected by biomass burning, biogenic VOCs, or other non- greater than 2 m/s, and 2) [CO] did not show a gradual in- urban pollutants. The 1[NOy]/1[CO], 1[benzene]/1[CO], crease of less than 100 ppbv/h in the afternoon as observed and 1[CH3CN]/1[CO] ratios were deduced from lin- on 12 March. On most other days, [CO] steadily decreased ear fits of the correlation graphs of 1-min data. The in the afternoon as the mixing depth grew above PTP. 1[NOy]/1[CO] ratio was 0.050 ppbv/ppbv (R = 0.99) on 11 We assume that concentrations of Ox and NOz were both and 12 March, 0.054 ppbv/ppbv (R=0.95) for the entire PTP vertically homogeneous within the mixing layer and horizon- dataset, and between 0.04 and 0.07 ppbv/ppbv at T0. The tally homogeneous within the ∼5.4 km air mass sampled dur- 1[benzene]/1[CO] ratio was 1.8 pptv/ppbv (R=0.93) on 11 ing this hour. From 12:15 to 13:15 we interpret the observed and 12 March 2006, 1.67 pptv/ppbv (R=0.87) for the entire increases in [Ox] and [NOz] to be mainly due to photochem- PTP dataset, and 1.1 to 1.8 at T0. The 1[CH3CN]/1[CO] istry, with an estimated uncertainty of 25% due to the coarse ratio on 12 March 2006 was between 0.2 to 0.5 pptv/ppbv. level of knowledge regarding the mixing layer depth and the Other measurements in urban air have been in the range 0.1– assumption of horizontal homogeneity for [Ox] and [NOz]. 0.3 (Kleinman et al., 2008; Knighton et al., 2007), while the Given these assumptions, the average Ox production rate ratios observed in air impacted by biomass burning are be- P(Ox) during this time is estimated using the observed time tween 1 and 7 (de Gouw et al., 2006). This suggests that rate of change of Ox (1[Ox]/1t; t ≡ time) and the calculated biomass burning probably did not have a significant impact Ox loss (destruction) rate L(Ox) according to Eq. (2): on the air on 12 March until 18:00, when a biomass burning PM plume was observed. Isoprene (canister) and other bio- 1[Ox]/1t=P(Ox) − L(Ox) (2) genic VOCs (e.g., terpenes) typically accounted for less than 10% of the calculated VOC reactivity over the entire PTP where the sign of L(Ox) is positive. The value of 1[Ox]/1t dataset, indicating there was only a small influence from bio- is 37 ppbv/h, calculated from the time series data. L(Ox) is genic emissions. calculated from R6–R7 and R8–R15 as summarized in Ta- ble 1: 3.2 Ox production and loss rates L(Ox) = 6 Rates ( R6, R7, R8 to R15) (3) Given the well-characterized, stagnant meteorology of 12 1 March 2006, the measurements are used to characterize nu- O3 + hν −→ O( D) + O2 (R8)

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2505

1 O( D) + H2O −→ 2OH (R9) (near urban emissions) and previously published literature values (Shorter et al., 2005). Net thermal of HO + O −→ OH + 2O (R10) 2 3 2 peroxyacyl nitrates would act as a source of Ox as well, but it is not evident from our measurements whether there was net OH + O −→ HO + O (R11) 3 2 2 formation or decomposition of peroxyacyl nitrates during the O3 + VOCs −→ products (R12) afternoon at PTP. We estimate that an upper limit to the con- tribution of peroxyacyl nitrate thermal decomposition to Ox NO2 + O3 −→ NO3 + O2 (R13) production during the stagnant period was 2 ppbv/h. The inferred P(Ox) value of 49±15 ppbv/h agrees within NO3 + hν −→ NO + O2 (R14) 50% with P(Ox) calculations based on the VOC reactiv- O3 dry deposition (R15) ity and estimated OH concentrations (Wood et al., 2009). Ozone production rates calculated using measurements of The NOz production rate is used to quantify Ox losses HO2 and NO and Eq. (1) reached even higher values of up from R6 and R7 because the products of these two reac- to 100 ppbv/h in the afternoon at more centrally-located ur- tions form observable NOz species. Because these two re- ban site in Mexico City (Dusanter et al., 2009; Shirley et al., actions account for approximately 80% of total NOz produc- 2006). These values are among the highest observed any- tion (based on the NOy speciation observed at the T1 site where in North America. A comparison of ozone produc- (Farmer et al., 2009), Ox loss by R6 and R7 is calculated tion rates in Phoenix, , and (Kleinman from the observed NOz production rate (Sect. 3.3.1) multi- et al., 2002) indicated peak ozone production rates of ap- plied by 0.8. The rates of R10 and R11 are calculated assum- proximately 5, 16, and 20 ppbv/h, respectively, with elevated ing an HO2 mixing ratio of 40 pptv and an OH rates of over 50 ppbv/h observed in petrochemical plumes of 6×106 cm−3 (as estimated later in this section in Houston. Peak instantaneous ozone production rates of and in section 3.3.3). measurements were not avail- 30–36 ppbv/h have been observed in Nashville (Daum et al., able on the afternoon of 12 March, and so the rate of alkene 2000; Thornton et al., 2002). The VOC reactivity at PTP cal- (R12) is based on measurements from other after- culated as 6kOH+VOC[VOC] was extremely high compared noons at PTP that had similar concentrations of CO, O3, and to the other locations – measurements from the afternoon of aromatic VOCs (which were measured continuously by the 11 and 12 March ranged from 13 s−1 to 20 s−1. In com- PTR-MS). Ox destruction via the minor NO-yielding branch parison, the 90th percentile VOC reactivities in Nashville, −1 −1 −1 of NO3 photolysis was calculated assuming that NO, NO2, Phoenix, and Houston were 6 s , 5 s , and 22 s , respec- and NO3 were in a photostationary state (Geyer et al., 2003). tively (Kleinman et al., 2002). The calculated VOC reac- Dry deposition of O3 was estimated using the boundary layer tivities at PTP are in rough agreement with measurements height of 0.9 km and a deposition velocity of 0.4 cm s−1 (We- of the total OH loss rate in Mexico City in 2003 (Shirley et sely and Hicks, 2000). al., 2006) of ∼25 s−1 in the afternoon – approximately 80% As seen in Table 1, the single greatest loss mechanism of of which was due to reaction with VOCs and CO (i.e., the −1 Ox was the conversion of NO2 into NOz compounds, which VOC reactivity was approximately 20 s ). A quantitative accounted for approximately 50% of Ox destruction. The discussion of the speciated VOC reactivity at PTP and T0 is importance of NO2 oxidation as an indirect O3 sink likely presented in Wood et al. (2009). decreased after one day of photochemistry, since most NOx Given the NO mixing ratio of 1.0±0.5 ppbv between 12:15 is converted to NOz on this time scale. The other Ox destruc- and 13:15, we estimate that the sum of [HO2] and [RO2] tion mechanisms are not expected to change as greatly after was 95±50 pptv using Eq. (1) and the rate constant of R4 one day. The relative importance of ozone photolysis (R8 (8.1×10−12 cm3 molecule−1 s−1) for oxidation of NO by and R9) as an Ox sink during the entire 11-day PTP data set RO2. Observations aboard the C-130 aircraft at comparable depended greatly on the meteorology, since the rate is pro- NOx concentrations (>10 ppbv) have shown that HO2 com- portional to the mixing ratio of vapor. prised slightly less than 50% of the sum of HO2 and RO2 With the value of 12 ppbv/h for L(Ox) as described (Cantrell et al., 2007). This value is useful for estimating the above, the Ox production rate is calculated using Eq. (3) as rate of the HO2 self-reaction and the rate of HO2+O3. The 49±15 ppbv/h, in which the uncertainty is based on the un- large uncertainties do not affect the conclusions of this paper. certainty in the fit of the slope of the time series data as well as uncertainty in the mixing layer depth. 3.3 Nitrogen oxides Primary emissions of NO2 are estimated to account for less than 1% of observed Ox at all times at PTP as esti- 3.3.1 Production rateof NOz mated by the quantity 0.04[NOy]/[Ox], where the average NO2/NOx emission ratio in Mexico City is assumed to be The value of P(NOz) can be inferred using a method 4%. This emission ratio is an upper limit based on early- similar to that used to infer P(Ox) in Sect. 3.2 (i.e. morning measurements of NO, NO2, and O3 at the T0 site 1[NOz]/1t=P(NOz)–L(NOz)). The value of 1[NOz]/1t www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2506 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City

1.0 [NO ]/[NO ] z y 1.0 0.8 [NO ]/[NO ] 3(PM1) z Entire PTP dataset (11 days) [HNO3]/[NOz] 0.6 ] 11 and 12 March 2006 z 0.8

Fraction 0.4 ])/[NO 0.2 (PM1)

- 0.6 0.0 3 35 10:00 AM 12:00 PM 2:00 PM 4:00 PM 6:00 PM NOz - 30 NO ] + [NO

3 (PM1) 3 0.4 HNO 25 3(g)

(ppb) NH3(g) 3

20 ([HNO 0.2 15 and NH z 10

NO 0.0 5 0.0 0.2 0.4 0.6 0.8 1.0 0 10:00 AM 12:00 PM 2:00 PM 4:00 PM 6:00 PM [NOz]/[NOy] 3/12/2006

Fig. 4. Speciation of NOz as a function of [NOz]/[NOy]. Mean val- Fig. 3. Partial speciation of NOy at PTP. The NOz data are de- − picted both as 3 minute averages (black squares) and as smoothed, ues of ([HNO3] + [NO3(PM1)])/ [NOz] using the entire PTP dataset 12 minute running averages (brown circles). Particulate nitrate and 11-12 March 2006 between 8:00 and 20:00 are displayed in − [NOz]/[NOy] bins of width 0.1 Error bars are ± one standard devi- (NO3(PM1)) accounts for 20-40% of NOz during the day, and ation of the 1-min averaged values. HNO3(g) accounts for less than 10%. Gaseous NH3 is included for comparison, though is not a part of NOy or NOz. − henceforth referred to as 6NO3 , accounted for 43% of NOz in the morning and decreased to a minimum value of 25% between 12:15 and 13:15, calculated as ([NOz]12:15– ± shortly before 15:00. [NOz]13:15)/1 h), was 6.8 1.7 ppbv/h. Losses of NOz during − this time period can be constrained by the observed decrease Since the NO3(PM1) measurement only accounts for accu- in 1[NOy]/1[CO], calculated as ([NOy]–0.6)/([CO]–130) mulation mode nitrate (vacuum aerodynamic diameter less (i.e., the background-adjusted [NOy]:[CO] ratio). Since CO than 1 µm), any uptake of NOy onto coarse-mode PM is not accounted for in this speciation. The dominant mode peak is relatively unreactive on a time scale of hours, it is useful as − a dilution tracer. During the stagnant hour, 1[NOy]/1[CO], of the NO3(PM1) size distribution was 330 nm at 10:00 and decreased by only 6%. If the overall urban [NOx]/[CO] emis- increased to 380 nm by 12:00. The same trend was observed sion ratio did not change enough to affect the ratio of cumu- in the ammonium aerosol size distributions. − lative (integrated) NOx and CO emitted, and if the decrease The partitioning of HNO3(g) and NO3(PM1) at other ground in [NOy]/[CO] was solely due to NOz losses rather than NOx sites in Mexico City has been described recently (Fountoukis losses, then this 6% decrease in [NOy]/[CO] would corre- et al., 2007; Hennigan et al., 2008; San Martini et al., 2006; spond to a 9% decrease in NOz, since [NOz]/[NOy] was Zheng et al., 2008). The high NH3 mixing ratios act to drive on average 0.65. The assumption regarding the changes in photochemically formed HNO3 into particulate ammonium the overall urban NOx/CO emission ratio is supported by the nitrate (NH4NO3). As the temperature increases in the af- 2006 emission inventory (CAM, 2008) and discussed further ternoon, NH4NO3 vaporizes, releasing HNO3 vapor and in- − in Sect. 3.3.2. Thus we calculate that P(NOz) is 9% higher creasing the ratio of HNO3(g) to NO3(PM1) (Fountoukis et al., than 1[NOz]/1t and equal to 7.4±1.9 ppbv/h. 2007; Hennigan et al., 2008; Zheng et al., 2008). − The relative partitioning of NOz among HNO3, NO3 , and 3.3.2 Speciation of NOy other compounds not measured at PTP (e.g., peroxyacyl ni- trates, alkyl nitrates, and coarse-mode nitrate PM) is deter- Between 12:15 and 13:15, [NOx]/[NOy] decreased from mined by the net production and loss rates of the main NOz − 0.5 to 0.2 due to the oxidation of NOx into NOz com- species (R6 – R7). Figure 4 shows that [6NO3 ]/[NOz] de- pounds (Fig. 3). The speciation of NOz can only partially creases with photochemical processing (i.e., with increasing be described since our directly measured individual NOz [NOz]/[NOy]). Figure 4 shows the means and standard de- species were restricted to accumulation-mode particulate ni- viations of the 1-minute averaged data from 12 March 2006 − trate (NO3(PM1)) and gas-phase HNO3. Particulate nitrate and the entire 11-day PTP dataset between 08:00 and 16:00, comprised 40% of NOz from 10:00 to 11:00 (Fig. 3) and which spans a range of meteorological conditions. This over- decreased to approximately 22% by 13:00. ac- all trend agrees with measurements aboard the C-130 air- − counted for less than 4% of NOz until 15:00, when it in- craft in which the sum of HNO3(g) and NO (as mea- − 3(PM1) creased to 12%. The sum of [HNO3(g)] and [NO3(PM1)], sured by an AMS) accounted for 30%–40% of NOz in aged

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2507 air masses ([NOz]/[NOy]>0.7) (Flocke, personal communi- trate) preclude a quantitative assessment of whether peroxy- cation, 2008). Speciated NOy measurements at the T1 site acyl nitrates were undergoing net formation or net decompo- exhibited similar trends (Farmer et al., 2009). sition during the stagnant hour. We note that the results of We discuss three possible factors contributing to the one modeling study (Lei et al., 2007) showed that there was − net formation of PAN compounds during the day in Mexico observed decrease in [6NO ]/[NOz]: 1) Deposition of 3 City, though not necessarily under the same conditions ob- HNO3(g), 2) A decrease in the role of HNO3 formation rel- served at PTP. Similarly, a box-model study of the Mexico ative to total NOz formation, and 3) a shift of the nitrate aerosol size distribution from the accumulation mode to the City outflow indicated that the ratio of the production rate of coarse mode (particles greater than 1 µm in diameter), which HNO3 to the net production rate of total PAN compounds de- is not detected by the AMS. creases in the early afternoon of the first day of photochem- istry (Madronich, 2006), though the temperature was held at The first possible explanation of the decrease in 10◦C in the model. The thermal decomposition lifetime (1/e) [6NO−]/[NO ] with photochemical age is enhanced depo- 3 z of PAN varies from ∼1 h at 24◦C to 2 days at 0◦C. sition of HNO relative to the other NO compounds. Gas- 3 z The third possible explanation for the observed decrease phase HNO has a high dry deposition velocity (Wesely and 3 in [6NO−]/[NO ] with increasing [NO ]/[NO ] is that there Hicks, 2000), but the fact that the high NH concentrations 3 z z y 3 was a shift in the size distribution of aerosol nitrate from the present in Mexico City convert most HNO into particulate 3 accumulation mode (where it can be detected by the AMS) (Hennigan et al., 2008) suggests that this to the coarse mode (Laskin et al., 2005). The AMS is most effect may not be as important as it is in low NH environ- 3 sensitive to particles with vacuum aerodynamic diameters ments. The extent to which HNO deposition caused the de- 3 between ∼60 nm and ∼800 nm (Canagaratna et al., 2007), crease in [6NO−]/[NO ] can be constrained by the small 3 z whereas the NO (and NO ) measurements do not have a decreases observed in 1[NO ]/1[CO] over the course of y z y particle size cut-off. This is a plausible explanation, espe- each day. Over the entire PTP dataset between 08:00 and cially since coarse-mode nitrate has been observed in Mex- 18:00, 1[NO ]/1[CO] decreased from a median value of y ico City during MILAGRO (Moffet et al., 2008; Querol et al., 0.050 at [NO ]/[NO ] values less than 0.2 (during the morn- z y 2008), however it cannot be quantified using the PTP dataset ing) down to 0.045 at [NO ]/[NO ] values greater than 0.6 z y as there was no measure of the coarse mode particulate ni- (which occurred on some but not all afternoons). If this de- trate mass. The decrease in the accumulation mode nitrate crease in [NOy]/[CO] is assumed to be due solely to depo- − mass during the morning is consistent with increasing am- sition of HNO3 and NO3(PM1), then the observed values of − bient temperature and the associated shift in the ammonium [6NO3 ]/[NOz] can be “corrected” upwards to account for nitrate equilibrium (Zheng et al., 2008). On 11 and 12 March, this deposition as described in Sect. 3.3.1. Such a correc- the AMS (and SMPS) mass distribution data indicated an in- tion is complicated by the fact that the NOx:CO emission crease in the particulate nitrate mass mode from 330 nm to ratio is not constant throughout the day. The key quantity 380 nm between 10:00 and 12:00. This shift in the aerosol that must be considered is the ratio of cumulative NOx emis- size distribution increased the fraction of total mass in parti- sions to cumulative CO emissions. The emissions inventory cles greater than 800 nm to which the AMS is less sensitive, (CAM, 2008) indicates that this ratio varies by less than 10% − which may also have contributed to the observations. − after 09:00. Overall, the value of [6NO3 ]/[NOz] can be in- In summary, the decreasing trend in [6NO3 ]/[NOz] with creased by at most a factor of 1.2 to account for HNO3 de- photochemical age displayed in Fig. 4 cannot be fully ex- position. This does not explain the overall decreasing trend − plained by deposition of HNO3(g), and was likely caused by in [6NO3 ]/[NOz] observed with increasing photochemical some combination of a decrease in the production rate of age. HNO3 relative to other NOz compounds, a shift of the ni- The second possible explanation for the decrease in trate aerosol mass distribution from the accumulation mode − [6NO3 ]/[NOz] over time is that there was a decrease in to the coarse mode, and a shift of the accumulation mode the HNO3 production rate relative to that of the other un- mass distribution toward particles greater than 800 nm (vac- measured NOz compounds, presumably peroxyacyl nitrates uum aerodynamic diameter). Similar conclusions regarding (R7) and multifunctional alkyl nitrates (R5b). The equilib- the importance of deposition and coarse-mode nitrate were rium constant for R7 for many organic peroxy radicals (such reached in Zheng et al. (2008). as the peroxyacetyl radical, CH3C(O)O2) is very sensitive to temperature, and the corresponding peroxyacyl nitrates can 3.3.3 Estimate of OH either be in a state of net formation or net thermal decom- − position depending on temperature and the concentrations of The increase in [6NO3 ] during the stagnant period can be − the relevant species. The lack of knowledge of the peroxy- used to infer the average OH concentration since 6NO3 is acyl nitrate mixing ratios themselves combined with the un- the result of R6 and [NO2] was measured. Between 12:15 − ± certainties in the mixing ratios of OH, NO, NO2, and the pre- and 13:15, [6NO3 ]/1t was 1.9 0.5 ppbv/h, which is ap- cursor compounds (e.g., acetaldehyde for peroxyacetyl ni- proximately 30%±8% of the total increase in NOz. Given www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2508 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City the average NO2 mixing ratio of 10 ppbv and the rate con- by reaction with NOx form observable NOz species (R5b– stant for R6, the average OH concentration during the stag- R7), P(ROx) is equal to P(NOz) when ozone production is nant period is estimated as 6×106 molecules cm−3, which is strongly VOC-limited (Eq. 3): in line with previous afternoon measurements of OH in Mex- ico City (Shirley et al., 2006). This value is an underestimate P(ROx) = L(ROx) ≈ P(NOz) (4) of [OH] if there was an increase in coarse-mode nitrate (and − − low by a factor of (HNO3(g)+NO3total)/(HNO3(g)+NO3PM1)). whereas P(NOz) is much less than P(ROx) when ozone Conversely, this is an overestimate of [OH] if HNO3 was not production is NOx-limited. Thus a comparison of P(ROx) the sole source of particulate nitrate (i.e., if there was signifi- and P(NOz) serves as a useful indicator of whether ozone cant partitioning of organic nitrates to the condensed phase). production is NOx-limited or VOC-limited. Note that Besides the comparison to previous measurements of OH in P(NOz)/P(ROx) is similar to the “LN /Q” quantity used by Mexico City, this estimate of [OH] is only used to quantify Kleinman et al. (2005). the rates of R11 and R19. The uncertainty in this estimate The validity of Eq. (3) assumes that all observable NOz does not affect the conclusions presented. For example, this compounds are formed by ROx-NOx reactions. The only ob- estimate would have to be more than a factor of 6 too low in servable NOz species that are not formed by ROx-NOx reac- order for R11 to account for an appreciable fraction (>10%) tions according to the current understanding of NOy chem- of the total Ox loss. istry are NO3,N2O5, and heterogeneously-formed HONO, and it is unlikely that any of these compounds accounts for 3.4 Radical budget and ozone production an appreciable fraction of NOz during the day. In order to compare P(ROx) and P(NOz), we calculate The photochemistry of ozone production and at- P(ROx) from the following ROx sources for the PTP data set: mospheric oxidation in general is driven by ROx reaction of O(1D) with (R8 and R9); photolysis (ROx≡OH+HO2+RO2) radicals. The design of effec- of the oxygenated VOCs formaldehyde, acetaldehyde, and tive air pollution abatement strategies requires knowledge acetone (R20–R22); and ozonolysis of alkenes (R23): of whether ozone production is “NOx-limited” or “VOC- limited” (NOx-saturated), which is intimately related to HCHO + hν + 2O2 −→ CO + 2HO2 (R20) reactions between ROx and NOx. At low NOx mixing ratios, self-reactions of ROx (R16–R19) dominate the removal CH3CHO + hν + 2O2 −→ CH3O2 + HO2 + CO (R21) processes of ROx, the Ox production rate increases with increasing NOx concentration, and Ox production is NOx- CH3COCH3 + hν + 2O2 −→ CH3C(O)O2 + CH3O2 (R22) limited. At higher NOx concentrations, reactions with NOx dominate the ROx removal processes, the Ox production O3+alkenes −→ products + RO2 + OH (R23) rate decreases with increasing NOx concentrations, and Ox production is VOC-limited. ROx yields for alkene-ozonolysis reactions are based on In order to investigate the radical budget and to determine the tabulated OH yields of Calvert et al. (2000) and are dou- whether ozone production was NOx-limited or VOC-limited bled to account for co-generated peroxy radicals. during the afternoon of 12 March, we compare the NOz pro- The time series of these photolytic ROx sources for 12 duction rate described in Sect. 3.3.1 to the production rate of March 2006 is depicted in Fig. 5. The average value of 1 ROx radicals. The reactions of ROx with NOx are the same P(ROx) from the reaction of O( D) with water vapor and the reactions that form NOz, i.e. formation of HNO3, RONO2, photolysis of oxygenated VOCs between 12:15 and 13:15 and RO2NO2 compounds (R5b, R6, R7). Thus the observed was 1.3±0.4 pptv/s. VOC canister measurements were not NOz production rate serves as an indicator of the sum of the made during this time period, and so the contribution of rates of R5b, R6, and R7. ROx sinks not accounted for by alkene ozonolysis to P(ROx) is not accurately known. This P(NOz) are from ROx self-reactions: quantity is estimated as 0.4±0.2 pptv/s, based on the ozonol- ysis rates calculated from other afternoon samples at PTP HO2 + HO2 −→ H2O2 + O2 (R16) with comparable values of concentrations of O3, CO, aro- matic VOCs, and photochemical age values calculated using RO2 + HO2 −→ ROOH + O2 (R17) a C3-benzene photochemical clock (Herndon et al., 2008). The total calculated P(ROx) is therefore 1.7±0.5 pptv/s, with RO2 + RO2 −→ products (R18) the range reflecting the uncertainty in the contribution from alkene ozonolysis and the uncertainty in the photolysis val- OH + HO2 −→ H2O + O2 (R19) ues. The observed value of P(NOz), 2.1±0.4 pptv/s, is a lower In steady state, the production rate of ROx (P(ROx)) limit to the true value of L(ROx) since non-NOx related equals the loss rate of ROx (L(ROx)). Since all losses of ROx ROx losses (R14-R17) do not form NOz. We estimate that

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2509 the water-assisted self reaction of HO2 accounts for ap- reactions. We note that this analysis of the calculated ROx proximately 0.1 pptv/s, based on an estimated HO2 mix- production rate and the inferred ROx loss rate does not pro- ing ratio of 40 pptv and the measured water vapor concen- vide information on the interconversions between OH, HO2, tration. The estimated rate of the reaction between RO2 and RO2; actual measurements of those species would be re- and HO2 using 40 pptv for both species and a rate constant quired for that analysis. −11 −3 −1 of 10 molecules cm s (Atkinson, 1994) is 0.3 pptv/s, In general, P(Ox) increases with increasing ROx produc- though there are strong indications that this rate constant is tion rates (Kleinman, 2005). The inferred value of P(Ox) high by up to an order of magnitude (Lelieveld et al., 2008; of 49 ppbv/h (13 pptv/s) for the observed P(ROx) rate of Thornton et al., 2002). Heterogeneous loss of HO2 is pos- 1.7±0.5 pptv/s also agrees well with that predicted by Lei sibly the most uncertain ROx loss process (Emmerson et al., et al. (2007). This agreement is perhaps fortuitous given the 2007; Thornton et al., 2008). uncertainties of both methods. The comparison of the calculated P(ROx) value P(ROx) exhibits an interesting asymmetry over the course (1.7±0.5 pptv/s) and L(ROx) inferred from P(NOz) of the day. Due to the elevated concentrations of oxygenated (2.1±0.4 pptv/s) indicates both that the budget of ROx VOCs in the morning, P(ROx) (excluding alkene ozonoly- sources and sinks is “closed” within the methodological sis) peaks at a value of 1.6 pptv/s at 11:00, whereas peak uncertainties (i.e., P(ROx) has been shown to equal L(ROx)) actinic flux and photolysis frequencies peak more than an and that ozone production is indeed VOC-limited since hour later. This overall temporal profile was observed dur- P(NOz) is roughly equal to L(ROx). That ozone production ing several but not all days at PTP. The peak and mean day- is VOC-limited is in agreement with predictions from time values of P(ROx) on 12 March were higher than on any previous photochemical models (Lei et al., 2007; Tie et al., other day at PTP between 8 March and 18 March. The mean 2007) and recent measurements (Nunnermacker et al., 2008; value of P(ROx) between 9:00 and 16:00 during this span Stephens et al., 2008). Although the large uncertainty bars of 11 days, under various meteorological conditions, was preclude a more quantitative insight into ROx sources, we 0.6 pptv/s, whereas on 12 March it was 1.2 pptv/s. note that it is not possible for P(NOz) to actually exceed Another interesting feature of ROx production on 12 P(ROx), assuming net formation of N2O5, NO3, and HONO March is that despite extremely high O3 mixing ratios was a negligible portion of P(NOz). Either P(NOz) has been (>150 ppbv), O3 photolysis accounted for at most 43% of overestimated or P(ROx) has been underestimated. The latter P(ROx) excluding alkene ozonolysis, similar to observations possibility is likely, since the calculation of P(ROx) does in the city center in 2003 (Volkamer et al., 2007). The rel- not include the net source of ROx from the photolysis of ative humidity was low on 12 March (less than 17% after HONO, H2O2, or oxygenated VOCs beyond formaldehyde, noon), limiting the rate of reaction between water vapor and acetaldehyde, and acetone. The HONO mixing ratio would O(1D) produced from the photolysis of ozone. The relative have to exceed the calculated photostationary state HONO importance of O3 photolysis as a ROx source increased later mixing ratio of ∼10 pptv by a factor of 10 to account for a in the month when the relative humidity increased greatly. net ROx source of 0.3 pptv/s. We estimate that photolysis The oxygenated VOCs (primarily HCHO) that account for of glyoxal (CHOCHO) contributes at most an additional the largest fraction of P(ROx) are both emitted directly in 0.1 pptv/s, based on the maximum glyoxal concentrations Mexico City at rates higher than in the US and are produced (∼1 ppbv) observed previously in Mexico City (Volkamer by VOC oxidation (Garcia et al., 2006; Kolb et al., 2004; et al., 2005). Studies of ROx production that have utilized Zavala et al., 2006). the Master Chemical Mechanism (Saunders et al., 2003) have invoked a large source of ROx from oxygenated VOCs 3.5 Efficiency of ozone production besides the three considered in our calculation of P(ROx) (Emmerson et al., 2007; Volkamer et al., 2007). It is not The efficiency of ozone production can be quantified with re- unreasonable that some combination of these sources could gard either to the NOx catalytic cycle or to the oxidation of account for an additional 0.2–0.3 pptv/s. It is unlikely that VOCs. The number of ozone molecules that are produced the reaction of electronically excited NO2 with water vapor per molecule of NOx before the NOx is removed from ac- acts as a significant ROx source during the afternoon due to tive photochemistry is known as the ozone production effi- the small solar zenith angles (Li et al., 2008). ciency (OPE) of NOx. Similarly, the amount of ozone that This use of NOz measurements as a tool for gleaning in- can be produced by the oxidation of individual VOCs has formation on ROx loss rates may be useful in smog cham- been studied in laboratory studies, yielding quantities such ber experiments as well as stagnant air masses. The alterna- as the ozone creation potential (Derwent et al., 1996) and the tive, more common method of quantifying ROx losses relies maximum incremental reactivity (Carter et al., 1995). We ex- on the explicit calculation of the rate of each ROxloss reac- ozone production in Mexico City using both the NOx- tion, which requires accurate knowledge of the reactants OH, based OPE metric and a new VOC-based metric that uses the HO2, and speciated RO2 mixing ratios as well as the rele- correlation of the 1[Ox]/1[CO] ratio with an aromatic-VOC vant rate constants, which are highly uncertain for many RO2 based photochemical clock. www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2510 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City

chemical and depositional losses of NOz (Nunnermacker et 2.5 P(NO ) al., 1998) and O . The OPE can also be difficult to interpret z S(O(1D), OVOCs) x P(RO )total x 1 if the observed air masses are inhomogeneous – i.e., if the O( D) + H2O n observed ozone is the result of multiple air masses of differ- 2.0 HCHO + h CH CHO + hn ent histories that have mixed prior to the observations (Liang 3 CH COCH + hn 3 3 and Jacobson, 2000). 1.5 The role that thermally labile peroxyacyl nitrates (such as PAN) have as a reservoir of both ROx and NOx complicates the interpretation of the OPE. Net formation of peroxyacyl Rate (pptv/s) 1.0 nitrates acts as a temporary sink of NOx (and Ox). Subse- quent thermal decomposition of peroxyacyl nitrates releases the NO , allowing it to catalyze the production of more O 0.5 x 3 - quite possibly in chemical environments where OPE(Inst) is different than where the NOx was initially emitted. Thus 0.0 the number of ozone molecules produced from each NOx de- 6:00 AM 9:00 AM 12:00 PM 3:00 PM 6:00 PM 9:00 PM pends on what temporal and spatial scale is considered. Shon et al. (2008) and Nunnermacker et al. (2008) report OPE val- Fig. 5. Comparison of the NOz production rate with the ROx ues (inferred from 1[Ox]/1[NOz]) that range from 5 within production rate during the afternoon of 12 March. ROx produc- the boundary layer up to 8.5 in the marine free troposphere tion rates from the photolysis of O3, HCHO, CH3CHO, and ace- tone are depicted as lines, with the sum depicted by the brown outside of Mexico City. The timing of NOx emissions can line (OVOC = oxygenated volatile ). The total also affect OPE values, since at night NOx emissions can lead P(ROx) value (black square) is calculated as the sum of the pho- to net destruction of Ox through reactions involving NO3 and tolytic ROx sources and an estimated 0.4 pptv/s from alkene ozonol- N2O5 (Brown et al., 2006). The OPE values determined be- ysis. The red square demarks the average NOz production rate ob- low in Sect. 3.5.2 and 3.5.3 are focused on daytime chemistry served between 12:15 and 13:15. on a time scale of less than 10 h. Between 12:15 and 13:15 on 12 March 2006, the value of P(Ox)/P(NOz) using the values for P(Ox) and P(NOz) de- PTP is an excellent site for examining both of these ap- rived in Sect. 3.1 and 3.2 was 6.4. This value for the OPE proaches given the known meteorology, the lack of proxi- reflects the gross Ox produced per NOx. This value is greater mate emission sources, and (for one hour during the after- than the correlation between the Ox and NOz concentrations noon of 12 March 2006) the quantified production and loss during this hour (5.4), which is more a measure of the net rates of both O and NO . x z Ox produced per NOx, and is also affected by deposition of NOz. The correlation is higher when all data between 3.5.1 Ozone production efficiency of NOx 09:00 and 16:00 on both 12 March and 11 March are in- cluded. These two days experienced similar air masses and The OPE of NOx has been extensively discussed over the last a comparable range of photochemical ages (as determined 20 years (Lin et al., 1988; Nunnermacker et al., 2000; Wang by [NOz]/[NOy]). The slope of the graph of Ox versus NOz et al., 1998). The instantaneous OPE is usually defined as (Fig. 5) for these two days is 6.2±0.2. To infer the gross P(Ox)/P(NOz), whereas the integrated OPE for an observed Ox produced per NOx, the slope can be “corrected” to ac- air mass is usually inferred by the correlation between Ox count for losses of Ox. Additionally, the NOz measurements and NOz (i.e., 1[Ox]/1[NOz]). The instantaneous OPE of can be corrected to account for cumulative NOz losses (de- an air mass is expected to vary with time, and is affected by position). As discussed in Sect. 3.3.1, we can infer that be- the temporal evolution of the partitioning of ROx among OH, tween 12:15 and 13:15 on 12 March approximately 9% of HO2, and RO2 and the partitioning of NOx between NO and the NOz was removed from the system by a comparison of NO2. the 1[NOy]/1[CO] ratio with that expected based on the cu- Inferred and/or modeled OPE values range from 2–8 in ur- mulative emissions of NOx and CO. If the emission ratio of ban (high-NOx) settings and power plumes (Kleinman, NOx/CO were actually a constant value of 0.05 throughout 2000;Nunnermacker et al., 2000) up to 46 for the mean OPE the day, the “corrected” value of [NOz] (i.e., the value of of the southern hemisphere (Wang et al., 1998). The large [NOz] that would have been measured had there been no NOz range of values reflects the non-linear dependence of ozone deposition) can be expressed by Eq. (4): production on its chemical precursors. In general, the OPE [NOz(CORRECTED)] = 0.05 × ([CO] − 130) − [NOx] (5) increases with the VOC reactivity to NOx ratio and is high- est in chemical environments in which ozone production is where the NOy measurements are ignored and instead are NOx-limited. The interpretation of inferred OPE values us- simulated by the CO measurements and scaled by the ing observations of 1[Ox]/1[NOz] is complicated by photo- NOx/CO emission ratio.

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2511

300 0.35

0.30 10:00 AM 12:30 PM 3:00 PM 5:30 PM 250

0.25

200 0.20 [CO] ]/ x 150 [O 0.15 (ppb) x O 0.10 100 0.05

O = (6.2 ± 0.2) NO + (43.4 ± 2.4), R2 = 0.86 50 x z 0.00 2 0 20 40 60 80 100 120x109 Oy = (7.6 ± 0.2) NOz(CORRECTED) + (44.4 ± 2.8), R = 0.94 t[OH] (molecules cm-3 s) 0 0 5 10 15 20 25 30 35 Fig. 7. Correlation of 1[Ox]/1[CO] on 12 March 2006 with cumu- NO (ppb) z lative OH exposure (1t[OH]) as inferred by a C9-aromatic photo- chemical clock. Fig. 6. Inference of the short-term ozone production efficiency (OPE) at PTP on 11 and 12 March 2006 between 08:30 and 18:00. The red points use the raw Ox and NOz data. Correcting for losses oxidation. The contribution of this Ox loss process likely de- of both species produces the blue points (see text). creased in importance after the NOx/NOy decreased to lower values (∼0.1).

The total Ox lost is determined by the integrated Ox de- The similarity between the corrected slope and the uncor- struction mechanisms described in Sect. 3.2. We use the rected slope indicate that the OPE can be reasonably approx- quantity “Oy” to represent the total integrated Ox produc- imated during one day of photochemistry in Mexico City by tion. In Sect. 3.2 the quantity 0.8×P(NOz) was used to esti- the correlation of [Ox] with [NOz]. Slightly higher Ox/NOz mate total Ox loss rate via NO2 oxidation, which accounted correlations were observed elsewhere in the city – for exam- for approximately 50% of the total Ox loss between 12:15 ple, our measurements at Santa Ana during an ozone-south and 13:15. If this value (50%) does not vary greatly over meteorological episode (deFoy et al., 2006) on 24 March ± the course of the day, then total integrated Ox losses can be 2006 yielded an (uncorrected) OPE value of 7.5 0.3. The approximated by Eq. (5): values derived here (6–8) are within the range of 4 to 12 cal- culated by Lei et al. (Lei et al., 2007) for NOx mixing ratios [OY] = [Ox] + 1.6[NOz] (6) greater than 10 ppbv.

The true percentage undoubtedly varies over the course of 3.5.2 1[Ox]/1[CO] and a VOC-based photochemical the day, contributing to the uncertainty of this approximation. clock [Oy] is between 5% and 25% higher than [Ox]. Figure 6 depicts the correlations on 11 and 12 March 2006 Since CO is a relatively unreactive tracer of dilution, the between Ox and NOz, which has a slope of 6.2, and that be- quantity 1[Ox]/1[CO] can be interpreted as the dilution- tween Oy and NOz(CORRECTED), which has a higher slope adjusted measure of integrated ozone production, similar to of 7.6. In most other cases where inferred OPE values have the use of the ratio of organic aerosol to CO in studies of sec- been corrected for losses (Nunnermacker et al., 1998), the ondary organic aerosol formation (Kleinman et al., 2008). In corrected OPE have been lower than the uncorrected OPE, the MCMA, on-road vehicles are the largest source of CO by in contrast to the case here. This is because the correction far, accounting for over 99% of total CO emissions (CAM, for Ox losses is greater than the correction for NOz. This is 2008). VOCs are emitted by on-road vehicles (34% by mass) partially just a difference in approach: by using Oy instead of as well as area sources (46%) and industrial sources (17%), Ox, the approach used here reflects the integrated ozone pro- though the most reactive VOCs are emitted by on-road ve- duction rather than the net ozone production per molecule of hicles (CAM, 2008; Velasco et al., 2007). Thus CO and the NOx oxidized. In the air masses examined, NOy was mostly most important ozone-producing VOCs are co-emitted, and conserved; the maximum correction to NOz was 20%. For emissions of total VOCs and CO overlap spatially. The CO more highly aged air masses, the relative magnitude of the mixing ratio is in most cases proportional to the total emis- Ox and NOz corrections might be different. The oxidation sions of VOCs, though the proportionality constant (e.g., the of NO2 was the dominant loss of Ox in the young air masses VOC/CO emission ratio) depends greatly on the of the observed, which had undergone less than 10 h of atmospheric emission sources . www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2512 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City

The increase in 1[Ox]/1[CO] can be related to the cu- have not undergone photochemical conversion, when [NOx] mulative OH exposure 1t[OH] (photochemical age) of the and [NOy] are equal. As the air mass is photochemically observed air mass using a C9-aromatic photochemical clock processed, Ox is produced, NOx is converted into NOz, and (Herndon et al., 2008; Roberts et al., 1984). The photochem- there are fresh emissions of NOx. Ignoring the effects of Ox ical age we use is based on PTR-MS measurements of ben- and NOy loss/deposition, the value of 1[Ox]/1[NOy] should zene and the sum of C9-aromatic compounds, which include approach the integrated OPE value of 1[Ox]/1[NOz], since C9H12 and C8H8O isomers. The calculation is described in NOy consists primarily (>90%) of NOz in highly aged air. Herndon et al. (2008). Since the air masses observed con- On 12 March 2006, the value of 1[Ox]/1[NOy] steadily sisted of a combination of fresh emissions and aged emis- increased during the day and reached 7 by 17:00, which is sions and are best described by a distribution of photochem- comparable to the OPE inferred in Fig. 6. This is expected ical ages, the single values derived using this photochemical since [NOz] and [NOy] are close in magnitude then, and thus clock are meant only as an estimate of the average OH expo- 1[Ox]/1[NOy]≈1[Ox]/1[NOz]. sure of the air. Comparison of the correlation of 1[Ox]/1[CO] with Figure 7 depicts the correlation between 1[Ox]/1[CO] 1t[OH] observed in Mexico City to other locations should and 1t[OH] between 09:00 and 17:00 on 12 March 2006, be done with the caveat that the values are greatly affected by with 1[Ox]/1[CO] calculated as ([Ox]–45)/([CO]–130). the VOC/CO emission ratios. This metric may be most use- Background values of 45 ppbv for Ox and 130 ppbv for ful for comparing ozone production during different times of CO are based on observations at night, when PTP was day or under different meteorological conditions at locations well above the nocturnal boundary layer. As expected, when the VOC/CO ratio does not vary greatly. 1[Ox]/1[CO] increases with 1t[OH], and there is little scatter in the plot (R2>0.9). The slope of a linear fit 4 Conclusions −12 −1 (not shown) is 3.8×10 ppbv Ox ppbv CO (molecules OH cm−3 s)−1, though it appears that it may be higher for Measurements of several ambient gas-phase and particulate the afternoon than for the morning data. This is supported by species from a unique mountain-top site in Mexico City have the fact that 1[Ox]/1[CO] should equal zero when 1t[OH] been used to characterize the chemistry of ozone production, is zero. Observations on other days were similar, with most nitrogen oxides, and the ROx budget, with an emphasis on slopes within the range of 3.6×10−12 to 5.6×10−12 ppbv one hour of data during which the rise of the mixing depth −1 −3 −1 Ox ppbv CO (molecules OH cm s) . The correlation was slow. Overall, the observations agree well with the pre- was poor when the air masses observed were affected by a dictions of photochemical models (Lei et al., 2007; Tie et al., range of emission sources within the same day (e.g., biomass 2007). burning vs. urban emissions). The observed time rate of change in the mixing ratios Variation in the slopes by time of day could be related to of Ox and NOz combined with calculations of the Ox loss differences in ozone photochemistry between the morning rates during the case study period (12 March 2006) were and afternoon. Within the boundary layer, overall concen- used to infer an Ox production rate of 49 ppbv/h and an trations of NOx and VOCs are highest in the morning be- NOz production rate of 7.6 ppbv/h. The dominant loss cause of the shallow mixing depth (typically less than 500 m process for Ox was the oxidation of NO2 to NOz com- between 06:00 and 11:00). Additionally, the VOC pounds (e.g., HNO3); photolysis of O3 was a minor con- differs: it consists primarily of unoxidized, primary VOCs in tributor to Ox loss due to the low relative humidity. A − the morning and a mix of primary and secondary VOCs (e.g., decrease in ([HNO3(g)]+[NO3(PM1)])/[NOz] with increasing oxygenated VOCs) in the afternoon. Thus it is conceivable [NOz]/[NOy] was observed. Deposition of HNO3 was a mi- that for a given amount of OH exposure, the different pho- nor contributor to this trend, whereas a shift of the aerosol tochemical conditions could lead to differences in the slope nitrate size distribution from the accumulation mode to the of 1[Ox]/1[CO] vs. 1t[OH] (e.g., if secondary VOCs are coarse mode and possibly a decrease in the HNO3 produc- more efficient at producing ozone). Changes in the slope of tion rate relative to total NOz production are plausible ex- Fig. 7 could also simply be caused by perturbations to the planations. A comparison of ROx production rates and the 1[Ox]/1[CO] ratio and the inferred OH exposure (1t[OH]) observed increase in NOz during a period of stagnant air due to changes in the emission ratios of the aromatic VOCs indicates that Ox production was VOC-limited and that the and CO. ROx budget was closed (i.e., P(ROx)=L(ROx)) to within the A plot of 1[Ox]/1[NOy] vs. 1t[OH] (not shown) looks methodological uncertainties. similar to Fig. 7 since NOy was mostly conserved in the air The ozone production efficiency of NOx on a time scale of masses observed. The quantity 1[Ox]/1[NOy] is related one day was approximately 7 as inferred by the correlation to the OPE of NOx in that it expresses the average number of [Ox] to [NOz]. Corrections to this correlation that account of ozone molecules that have been produced per molecule for losses of Ox and NOz increase the inferred value of the of NOx emitted (rather than ozone produced per NOx oxi- ozone production efficiency by less than 30%. A new met- dized). This quantity is initially zero in fresh emissions that ric for assessing the efficiency of ozone production has been

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2513 proposed and is based on the correlation of the [Ox]:[CO] of ambient with the Aerodyne Aerosol Mass Spectrome- ratio with the average OH exposure as inferred by ratios of ter, Mass Spectrom. Rev., 26, 185–222, doi:10.1002/mas.20115, aromatic VOCs. 2007. Although this analysis focused on only a short time pe- Cantrell, C., Anderson, R., Mauldin, L., Kociuch, E., McCoy, J., riod, the chemistry characterized is in agreement with pre- Eisele, F., Apel, E., Reimer, D., Hills, A., Karl, T., Knapp, D., Montzka, D., Fried, A., Weibring, P., Walega, W., Richter, D., vious modeling work and lends confidence to the state of Flocke, F., Zheng, W., Emmons, L., Crawford, J., Dunlea, E., knowledge regarding ozone chemistry, nitrogen oxides, and DeCarlo, P., Clarke, T., Howell, S., Cohen, R., Shetter, R., Hall, the underlying fast radical chemistry. A few aspects of the S., Russell, L., Kok, G., Weber, R., Madronich, S., Wennberg, chemistry, however, have not been commonly observed in P., Crounse, J., McCabe, D., and Holloway, J.: HOx Behavor other locations. For example, gas-phase HNO3 comprised a as Observed Aboard the C-130 during MILAGRO, MILAGRO small fraction of NOz (less than 15% usually), whereas par- Science Meeting, Mexico City, 2007. ticulate nitrate comprised a much larger fraction - mainly Carslaw, N., Creasey, D. J., Harrison, D., Heard, D. E., Hunter, M. due to the high concentrations of NH3. The photolysis C., Jacobs, P. J., Jenkin, M. E., Lee, J. D., Lewis, A. C., Pilling, of ozone and subsequent reaction of O(1D) with water va- M. J., Saunders, S. M., and Seakins, P. W.: OH and HO2 radical chemistry in a forested region of north-western Greece, Atmos. por was a minor destruction channel for O3 during the first Environ., 35, 4725–4737, 2001. ∼24 h of , and even during high O3 Carter, W. P. L., Pierce, J. A., Luo, D., and Malkina, I. L.: En- events ([O ]>150 ppbv) ozone photolysis accounted for less 3 vironmental chamber study of maximum incremental reactivities than half of total ROx sources in dry air masses. Although of volatile organic compounds, Atmos. Environ., 29, 2499–2511, rare in the literature, these characteristics are not necessarily 1995. unique considering that the atmosphere above most develop- Daum, P. H., Kleinman, L. I., Imre, D., Nunnermacker, L. J., Lee, Y. ing megacities remains largely uncharacterized. N., Springston, S. R., Newman, L., Weinstein-Lloyd, J., Valente, R. J., Imhoff, R. E., Tanner, R. L., and Meagher, J. F.: Analysis of Acknowledgements. This work was funded in part by NSF O3 formation during a stagnation episode in central Tennessee in grants (ATM-528227 & ATM-0528170) and DOE grants (DE- summer 1995, J. Geophys. Res.-Atmos., 105, 9107–9119, 2000. FGO2-05ER63982 & DE-FGO2-05ER63980). We gratefully de Foy, B., Caetano, E., Magana,˜ V., Zitacuaro,´ A., Cardenas,´ acknowledge Luisa Molina and Linsey Marr for providing the two B., Retama, A., Ramos, R., Molina, L. T., and Molina, M. chemiluminescence analyzers. We thank Manuel Quinones˜ of J.: Mexico City basin wind circulation during the MCMA- Televisa for providing power and on-site support at PTP, Rafael 2003 field campaign, Atmos. Chem. Phys., 5, 2267–2288, 2005, Ramos for logistical assistance, Sasha Madronich of NCAR for http://www.atmos-chem-phys.net/5/2267/2005/. assistance with the TUV model, and Delphine Farmer for helpful de Foy, B., Varela, J. R., Molina, L. T., and Molina, M. J.: conversations. The Environmental Protection Agency Rapid ventilation of the Mexico City basin and regional fate through its Office of Research and Development collaborated in the of the urban plume, Atmos. Chem. Phys., 6, 2321–2335, 2006, research described here. It has been subjected to Agency review http://www.atmos-chem-phys.net/6/2321/2006/. and approved for publication. de Foy, B., Fast, J. D., Paech, S. J., Phillips, D., Walters, J. T., Coulter, R. L., Martin, T. J., Pekour, M. S., Shaw, W. J., Edited by: S. Madronich Kastendeuch, P. P., Marley, N. A., Retama, A., and Molina, L. T.: Basin-scale wind transport during the MILAGRO field campaign and comparison to climatology using cluster analysis, References Atmos. Chem. Phys., 8, 1209–1224, 2008, http://www.atmos- chem-phys.net/8/1209/2008/. Atkinson, R.: Gas-phase tropospheric chemistry of organic com- de Gouw, J. A., Middlebrook, A. M., Warneke, C., Goldan, P. pounds, Journal of Physical Chemistry Ref. Data, M2, 1994. D., Kuster, W. C., Roberts, J. M., Fehsenfeld, F. C., Worsnop, Brown, S. S., Neuman, J. A., Ryerson, T. B., Trainer, M., Dube,´ W. D. R., Canagaratna, M. R., Pszenny, A. A. P., Keene, W. C., P., Holloway, J. S., Warneke, C., De Gouw, J. A., Donnelly, S. G., Marchewka, M., Bertman, S. B., and Bates, T. S.: Budget of and Atlas, E.: Nocturnal odd-oxygen budget and its implications organic in a polluted atmosphere: Results from the New for ozone loss in the lower troposphere, Geophys. Res. Lett., 33, England Air Quality Study in 2002, J. Geophys. Res.-Atmos., doi:10.1029/2006GL025900, 2006. 110, doi:10.1029/2004JD005623, 2005. Calvert, J. G., Atkinson, R., Kerr, J. A., Madronich, S., Moortgat, de Gouw, J. A., Warneke, C., Stohl, A., Wollny, A. G., Brock, G. K., Wallington, T. J., and Yarwood, G.: The Mechanisms of C. A., Cooper, O. R., Holloway, J. S., Trainer, M., Fehsenfeld, Atmospheric Oxidation of the Alkenes, Oxford University Press, F. C., Atlas, E. L., Donnelly, S. G., Stroud, V., and Lueb, A.: Oxford, 277–278, 2000. Volatile organic compound composition of merged and aged for- CAM: Inventario de emisiones 2006 de la Zona Metropolitana del est fires from Alaska and Western Canada, J. Geophys. Res., 111, Valle de Mexico, http://www.sma.df.gob.mx, 2008. doi:10.1029/2005JD006175, 2006. Canagaratna, M. R., Jayne, J. T., Jimenez, J. L., Allan, J. D., Al- deFoy, B., Varela, J. R., Molina, L. T., and Molina, M. J.: farra, M. R., Zhang, Q., Onasch, T. B., Drewnick, F., Coe, H., Rapid ventilation of the Mexico City basin and regional fate Middlebrook, A., Delia, A., Williams, L. R., Trimborn, A. M., of the urban plume, Atmos. Chem. Phys., 6, 2321–2335, 2006, Northway, M. J., DeCarlo, P. F., Kolb, C. E., Davidovits, P., and http://www.atmos-chem-phys.net/6/2321/2006/. Worsnop, D. R.: Chemical and microphysical characaterization www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2514 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City

Derwent, R. G., Jenkin, M. E., and Saunders, S. M.: Photochemical ico City, Geophys. Res. Lett., 35, doi:10.1029/2008GL034058, ozone creation potentials for a large number of reactive hydrocar- 2008. bons under European conditions, Atmos. Environ., 30, 181–199, Herndon, S. C., Jayne, J. T., Zahniser, M. S., Worsnop, D. R., 1996. Knighton, B., Alwine, E., Lamb, B. K., Zavala, M., Nelson, Drewnick, F., Hings, S. S., DeCarlo, P., Jayne, J. T., Gonin, M., D. D., and McManus, J. B.: Characterization of urban pollutant Fuhrer, K., Weimer, S., Jimenez, J. L., Demerjian, K. L., Bor- emission fluxes and ambient concentration distributions using a rman, S., and Worsnop, D. R.: A New Time-of-Flight Aerosol mobile laboratory with rapid response instrumentation, Faraday Mass Spectrometer (TOF-AMS) – Instrument Description and Discuss, 130, 327–339, 2005. First Field Deployment, Aerosol Sci. Technol., 39, 637–658, Herndon, S. C., M. S. Zahniser, D. D. Nelson Jr., J. Shorter, 2005. J. B. McManus, R. Jimenez,´ C. Warneke, and Gouw, J. A. Dusanter, S., Vimal, D., Stevens, P. S., Volkamer, R., and Molina, d.: Airborne measurements of HCHO and HCOOH during the L. T.: Measurements of OH and HO2 concentrations during the New England Air Quality Study 2004 using a pulsed quan- MCMA-2006 field campaign Part 1: Deployment of the Indi- tum cascade laser spectrometer, J. Geophys. Res., 112, D10S03, ana University laser-induced fluorescence instrument, Atmos. doi:10.1029/2006JD007600., 2007. Chem. Phys., 9, 1665–1685, 2009, http://www.atmos-chem- Khan, H. A.: Benzene’s : a consolidated short review of phys.net/9/1665/2009/. human and animal studies, Human Exp. Tox., 26, 677–685, 2007. Emmerson, K. M., Carslaw, N., Carslaw, D. C., Lee, J. D., McFig- Kleinman, L. I.: Ozone process insights from field experiments – gans, G., Bloss, W. J., Gravestock, T., Heard, D. E., Hopkins, J., part II: Observation-based analysis for ozone production, Atmos. Ingham, T., Pilling, M. J., Smith, S. C., Jacob, M., and Monks, P. Environ., 34, 2023–2033, 2000. S.: Free radical modelling studies during the UK TORCH Cam- Kleinman, L. I., Daum, P. H., Imre, D., Lee, Y. N., Nunnermacker, paign in Summer 2003, Atmos. Chem. Phys., 7, 167–181, 2007, L. J., Springston, S. R., Weinstein-Lloyd, J., and Rudolph, J.: http://www.atmos-chem-phys.net/7/167/2007/. Ozone production rate and reactivity in 5 urban ar- Farmer, D. K., Cohen, R. C., Perring, A. E., et al.: NOy partitioning eas: A cause of high ozone concentration in Houston, Geophys. and the role of alkyl nitrates in air quality in the Mexico City Res. Lett., 29(10), 1467, doi:10.1029/2001GL014569, 2002. area, Atmos. Chem. Phys., in preparation, 2009. Kleinman, L. I.: The dependence of tropospheric ozone production Fast, J. D., De Foy, B., Acevedo Rosas, F., Caetano, E., Carmichael, rate on ozone precursors, Atmos. Environ., 39, 575–586, 2005. G., Emmons, L., McKenna, D., Mena, M., Skamarock, W., Kleinman, L. I., Springston, S. R., Daum, P. H., Lee, Y. N., Nun- Tie, X., Coulter, R. L., Barnard, J. C., Wiedinmyer, C., and nermacker, L. J., Senum, G. I., Wang, J., Weinstein-Lloyd, J., Madronich, S.: A meteorological overview of the MILAGRO Alexander, M. L., Hubbe, J., Ortega, J., CAnagaratna, M. R., field campaigns, Atmos. Chem. Phys., 7, 2233–2257, 2007, and Jayne, J.: The time evolution of aerosol composition over http://www.atmos-chem-phys.net/7/2233/2007/. the Mexico City plateau, Atmos. Chem. Phys., 8, 1559–1575, Fountoukis, C., Nenes, A., Sullivan, A., Weber, R., VanReken, T., 2008, http://www.atmos-chem-phys.net/8/1559/2008/. Fischer, M., Matias, E., Moya, M., Farmer, D., and Cohen, R. Knighton, W. B., Fortner, E. C., Herndon, S. C., Wood, E. C., Cana- C.: Thermodynamic characterization of Mexico City aerosol dur- garatna, M., Jayne, J., Kroll, J. H., Onasch, T., Trimborn, A., ing MILAGRO 2006, Atmos. Chem. Phys., 9, 2141–2156, 2009, Worsnop, D., and Kolb, C.: Examination of Biomass Burning http://www.atmos-chem-phys.net/9/2141/2009/. Tracer Signatures in Urban Environments, EOS, Transactions, Garcia, A. R., Volkamer, R., Molina, L. T., Molina, M. J., Samuel- American Geophysical Union, A33D-1569, 2007. son, J., Mellqvist, J., Galle, B., Herndon, S. C., and Kolb, Kolb, C. E., Herndon, S. C., McManus, J. B., Shorter, J. H., Zah- C. E.: Separation of emitted and photochemical formaldehyde niser, M. S., Nelson, D. D., Jayne, J. T., Canagaratna, M. R., and in Mexico City using a statistical analysis and a new pair of Worsnop, D. R.: Mobile Laboratory with Rapid Response In- gas-phase tracers, Atmos. Chem. Phys., 6, 4545–4557, 2006, struments for Real-Time Measurements of Urban and Regional http://www.atmos-chem-phys.net/6/4545/2006/. and Particulate Distributions and Emission Source Geyer, A., Alicke, B., Ackermann, R., Martinez, M., Harder, H., Characteristics, Environ. Sci. Technol., 38, 5694–5703, 2004. Brune, W., Di Carlo, P., Williams, E., Jobson, T., Hall, S., Shet- Laskin, A., Wietsma, T. W., Krueger, B. J., and Grassian, V. ter, R., and Stutz, J.: Direct observations of daytime NO3: Im- H.: Heterogenous chemistry of individual dust par- plications for urban boundary layer chemistry, J. Geophys. Res., ticles with nitric acid: A combined CCSEM/EDX, ESEM, 108, 4368, doi:4310.1029/2002JD002967, 2003. and ICP-MS study, J. Geophys. Res.-Atmos., 110(D10), Hennigan, C. J., Sullivan, A. P., Fountoukis, C. I., Nenes, A., doi:10.1029/2004JD005206, 2005. Hecobian, A., Vargas, O., Peltier, R. E., Case Hanks, A. T., Lei, W., de Foy, B., Zavala, M., Volkamer, R., and Molina, L. T.: Huey, L. G., Lefer, B. L., Russell, A. G., and Weber, R. J.: Characterizing ozone production in the Mexico City Metropoli- On the volatility and production mechanims of newly formed ni- tan Area: a case study using a chemical transport model, Atmos. trate and water soluble organic aerosol in Mexico City, Atmos. Chem. Phys., 7, 1347–1366, 2007, http://www.atmos-chem- Chem. Phys., 8, 3761–3768, 2008, http://www.atmos-chem- phys.net/7/1347/2007/. phys.net/8/3761/2008/. Lelieveld, J., Butler, T. M., Crowley, J. N., Dillon, T. J., Fischer, Herndon, S., Onasch, T., Wood, E. C., Kroll, J. H., Canagaratna, M., H., Ganzeveld, L., Harder, H., Lawrence, M. G., Martinez, M., Jayne, J., Zavala, M., Knighton, W. B., Mazzoleni, C., Dubey, M. Taraborrelli, D., and Williams, J.: Atmospheric oxidation capac- K., Ulbrich, I., Jimenez, J. L., Seila, R., de Gouw, J. A., De Foy, ity sustained by a tropical forest, Nature, 452, 737–740, 2008. B., Fast, J., Molina, L., Kolb, C. E., and Worsnop, D. R.: The cor- Li, S., Matthews, J., and Sinha, A.: Atmospheric relation of secondary organic aerosol with odd oxygen in Mex- Production from Electronically Excited NO2 and H2O, Science,

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/ E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City 2515

319, 1657–1660, 2008. plant plumes: Analysis of field data, J. Geophys. Res.-Atmos., Liang, J. and Jacobson, M. Z.: Effects of subgrid segregation on 105, 9165–9176, 2000. ozone production efficiency in a chemical model, Atmos. Envi- Nunnermacker, L. J., Weinstein-Lloyd, J. B., Hillery, B., Giebel, B., ron., 34, 2975–2982, 2000. Kleinman, L. I., Springston, S. R., Daum, P. H., Gaffney, J., Mar- Lim, Y. B. and Ziemann, P. J.: Products and mechanism of sec- ley, N., and Huey, G.: Aircraft and ground-based measurements ondary organic aerosol formation from reactions of n- of during the 2006 MILAGRO field campaign, with OH radicals in the presence of NOx, Environ. Sci. Technol., Atmos. Chem. Phys., 8, 7619–7636, 2008, http://www.atmos- 39, 9229–9236, 2005. chem-phys.net/8/7619/2008/. Lin, X., Trainer, M., and Liu, S. C.: On the Nonlinearity of the Querol, X., Pey, J., Minguillon, M. C., Perez, N., Alastuey, A., Tropospheric Ozone Production, J. Geophys. Res., 93, 15879– Viana, M., Moreno, T., Bernabe, R. M., Blanco, S., Carde- 15888, 1988. nas, B., Vega, E., Sosa, G., Escalona, S., Ruiz, H., and Arti- Lindinger, W., Hansel, A., and Jordan, A.: Proton-transfer-reaction nano, B.: PM speciation and sources in Mexico during the (PTR-MS): on-line monitoring of volatile or- MILAGRO-2006 Campaign, Atmos. Chem. Phys., 8, 111–128, ganic compounds at pptv levels, Chemical Society Reviews, 27, 2008, http://www.atmos-chem-phys.net/8/111/2008/. 347, 1998. Roberts, J. M., Fehsenfeld, F. C., Liu, S. C., Bollinger, M. J., Hahn, Liu, P. S. K., Deng, R., Smith, K. A., Williams, L. R., Jayne, J. C., Albritton, D. L., and Sievers, R. E.: Measurements of aro- T., Canagaratna, M. R., Moore, K., Onasch, T. B., Worsnop, D. matic hydrocarbon ratios and NOx concentrations in the rural R., and Deshler, T.: Transmission Efficiency of an Aerodynamic troposphere: Estimates of air mass photochemical age and NOx Focusing Lens System: Comparison of Model Calculations and removal rate, Atmos. Environ., 18, 2421–2432, 1984. Laboratory Measurements for the Aerodyne Aerosol Mass Spec- Rogers, T. M., Grimsrud, E. P., Herndon, S. C., Jayne, J. T., Kolb, trometer, Aerosol Sci. Technol., 41, 721–733, 2007. C. E., Allwine, E., Westberg, H., Lamb, B. K., Zavala, M., Madronich, S.: Chemical Evolution of gaseous air pollutants down- Molina, L. T., Molina, M. J., and Knighton, W. B.: On-road wind of tropical megacities: Mexico City case study, Atmos. En- measurements of volatile organic compounds in the Mexico City viron., 40, 6012–6018, 2006. metropolitan area using proton transfer reaction mass spectrom- Matthew, B. M., Middlebrook, A. M., and Onasch, T. B.: Collec- etry, Int. J. Mass Spectrom., 252, 26–37, 2006. tion Efficiencies in an Aerodyne Aerosol Mass Spectrometer as Ryerson, T. B., Trainer, M., Holloway, J. S., Parrish, D. D., Huey, a Function of Particle Phase for Laboratory Generated Aerosols, L. G., Sueper, D. T., Frost, G. J., Donnelly, S. G., Schauffler, S., Aerosol Sci. Technol., 42, 884–898, 2008. Atlas, E. L., Kuster, W. C., Goldan, P. D., Hubler, G., Meagher, McCreanor, J., Cullinan, P., Nieuwenhuijsen, M. J., Stewart-Evans, J. F., and Fehsenfeld, F. C.: Observations of ozone formation in J., Malliarou, E., Jarup, L., Harrington, R., Svartengren, M., power plant plumes and implications for ozone control strategies, Han, I., Ohman-Strickland, P., Chung, K. F., and Zhang, J.: Res- Science, 292, 719–723, 2001. piratory Effects of Exposure to Diesel Traffic in Persons with Salcedo, D., Onasch, T. B., Dzepina, K., Canagaratna, M. R., , The New England Journal of Medicine, 357, 2348- Zhang, Q., Huffman, J. A., DeCarlo, P. F., Jayne, J. T., Mortimer, 2358, 2007. P., Worsnop, D. R., Kolb, C. E., Johnson, K. S., Zuberi, B., Marr, Moffet, R. C., de Foy, B., Molina, L. T., Molina, M. J., and L. C., Volkamer, R. M., Molina, L. T., Molina, M. J., Carde- Prather, K. A.: Measurement of ambient aerosols in north- nas, B., Bernabe, R. M., Marquez, C., Gaffney, J. S., Marley, N. ern Mexico City by single particle mass spectrometry, Atmos. A., Laskin, A., Shutthanandan, V., Xie, Y., Brune, W., Lesher, Chem. Phys., 8, 4499–4516, 2008, http://www.atmos-chem- R., Shirley, T., and Jimenez, J. L.: Characterization of ambient phys.net/8/4499/2008/. aerosols in Mexico City during the MCMA-2003 campaign with Molina, L. T., Molina, M. J., Slott, R. S., Kolb, C. E., Gbor, P. K., aerosol mass spectrometry: results from the CENICA Supersite., Meng, F., Singh, R. B., Galvez, O., Sloan, J. J., Anderson, W. Atmos. Chem. Phys., 6, 925–946, www.atmos-chem-phys.net/6/ P., Tang, X., Hu, M., Xie, S., Shao, M., Zhu, T., Zhang, Y. H., 925/2006/, 2006. Gurjar, B. R., Artaxo, P. E., Oyola, P., Gramsch, E., Hidalgo, D., San Martini, F. M., Dunlea, E. J., Grutter, M., Onasch, T. B., Jayne, and Gertler, A. W.: Air Quality in Selected Megacities, J. Air J. T., Canagaratna, M. R., Worsnop, D. R., Kolb, C. E., Shorter, Waste Manage. Assoc., 55, 1–73, 2004. J. H., Herndon, S. C., Zahniser, M. S., Ortega, J. M., McRae, Nelson, D. D., McManus, B., Urbanski, S., Herndon, S., and Zah- G. J., Molina, L. T., and Molina, M. J.: Implementation of a niser, M. S.: High precision measurements of atmospheric ni- Markov Chain Monte Carlo method to inorganic aerosol mod- trous oxide and using thermoelectrically cooled mid- eling of observations from the MCMA-2003 campaign - Part I: infrared quantum cascade lasers and detectors, Spectroc. Acta Model description and application to the La Merced site, Atmos. Pt. A-Molec. Biomolec. Spectrom., 60, 3325–3335, 2004. Chem. Phys, 6, 4867–4888, 2006. Nunnermacker, L. J., Imre, D., Daum, P. H., Kleinman, L., Lee, Sander, S. P., Friedl, R. R., Golden, D. M., Kurylo, M. J., Moortgat, Y. N., Springston, S. R., Newman, L., Weinstein-Lloyd, J., Luke, G. K., Wine, P. H., Ravishankara, A. R., Kolb, C. E., Molina, M. W. T., Banta, R., Alvarez, R., Senff, C., Sillman, S., Holdren, M., J., Finlayson-Pitts, B. J., Huie, R. E., and Orkin, V. L.: Chem- Keigley, G. W., and Zhou, X.: Characterization of the Nashville ical Kinetics and Photochemical Data for Use in Atmospheric urban plume on July 3 and July 18, 1995, J. Geophys. Res., 103, Studies: Evaluation Number 15, Jet Propulsion LaboratoryJPL 28129–28148, 1998. Publication 06-2, 2006. Nunnermacker, L. J., Kleinman, L. I., Imre, D., Daum, P. H., Lee, Saunders, S. M., Jenkin, M. E., Derwent, R. G., and Pilling, M. Y. N., Lee, J. H., Springston, S. R., Newman, L., and Gillani, N.: J.: Protocol for the development of the Master Chemical Mech- NOy lifetimes and O3 production efficiencies in urban and power anism, MCM v3 (Part A): tropospheric degradation of non- www.atmos-chem-phys.net/9/2499/2009/ Atmos. Chem. Phys., 9, 2499–2517, 2009 2516 E. C. Wood et al.: Ozone, NOx, and radicals in Mexico City

aromatic volatile organic compounds, Atmos. Chem. Phys., 3, Kolb, C. E., Zavala, M., De Foy, B., Volkamer, R., Molina, L. T., 161–180, 2003, http://www.atmos-chem-phys.net/3/161/2003/. and Molina, M. J.: Distribution, magnitudes, reactivities, ratios Seila, R. L., Main, H. H., Arriaga, J. L., Martinez, G., and Ramadan, and diurnal patterns of volatile organic compounds in the Valley A. B.: Atmospheric volatile organic compound measurements of Mexico during the MCMA 2002 & 2003 field campaigns, At- during the 1996 Paso del Norte Study, The Science of the Total mos. Chem. Phys., 7, 329–353, 2007, http://www.atmos-chem- Environment, 276, 153–169, 2001. phys.net/7/329/2007/. Shaw, W. J., Pekour, M. S., Coulter, R. L., Martin, T. J., and Wal- Volkamer, R., Molina, L. T., Molina, M. J., Shirley, T., and Brune, ters, J. T.: The daytime mixing layer observed by , W. H.: DOAS measurement of glyoxal as an indicator for fast profiler and during MILAGRO, Atmos. Chem. Phys. Dis- VOC chemistry in urban air, Geophys. Res. Lett., 32, L08806,, cuss., 15025–15065, 2007. doi:10.1029/2005GL022616, 2005. Shirley, T. R., Brune, W. H., Ren, X., Mao, J., Lesher, R., Car- Volkamer, R., Jimenez, J. L., San Martini, F., Dzepina, K., Zhang, denas, B., Volkamer, R., Molina, L. T., Molina, M. J., Lamb, Q., Salcedo, D., Molina, L. T., Worsnop, D. R., and Molina, M. B., Velasco, E., Jobson, T., and Alexander, M.: Atmospheric J.: Secondary organic aerosol formation from anthropogenic air oxidation in the Mexico City Metropolitan Area (MCMA) dur- pollution: Rapid and higher than expected, Geophys. Res. Lett., ing April 2003, Atmos. Chem. Phys., 6, 2753-2765, 2006, 33, L17811, doi:10.1029/2006GL026899, 2006. http://www.atmos-chem-phys.net/6/2753/2006/. Volkamer, R., Sheehy, P., Molina, L., and Molina, M.: Oxidative Shon, Z. H., Madronich, S., Song, S. K., Flocke, F. M., Knapp, D. capacity of the Mexico City atmosphere –Part 1: A radical source J., Anderson, R. S., Shetter, R. E., Cantrell, C. A., Hall, S. R., perspective, Atmos. Chem. Phys. Discuss., 7, 5365–5412, 2007, and Tie, X.: Characteristics of the NO-NO2-O3 system in differ- http://www.atmos-chem-phys-discuss.net/7/5365/2007/. ent chemical regimes during the MIRAGE-Mex field campaign, Wang, Y. H., Logan, J. A., and Jacob, D. J.: Global simulation Atmos. Chem. Phys., 8, 7153–7164, 2008, http://www.atmos- of tropospheric O3-NOx-hydrocarbon chemistry 2. Model eval- chem-phys.net/8/7153/2008/. uation and global ozone budget, J. Geophys. Res.-, Shorter, J. H., Herndon, S., Zahniser, M. S., Nelson, D. D., 103, 10727–10755, 1998. Wormhoudt, J., Demerjian, K. L., and Kolb, C. E.: Real-time Wesely, M. L. and Hicks, B. B.: A review of the current status of Measurements of Emissions from In-use New knowledge on dry deposition, Atmos. Environ., 34, 2261–2282, York City Transit Buses using a Chase Vehicle, Environ. Sci. 2000. Technol, 39, 7991–8000, 2005. Williams, E. J., Baumann, K., Roberts, J. M., Bertman, S. B., Nor- Stephens, S., Madronich, S., Wu, F., Olson, J. B., Ramos, R., ton, R. B., Fehsenfeld, F. C., Springston, S. R., Nunnermacker, Retama, A., and Munoz, R.: Weekly patterns of Mexico L. J., Newman, L., Olszyna, K., Meagher, J., Hartsell, B., Edger- City’s surface concentrations of CO, NOx, PM10 and O3 dur- ton, E., Pearson, J. R., and Rodgers, M. O.: Intercomparison of ing 1986–2007, Atmos. Chem. Phys., 8, 5313–5325, 2008, ground-based NOy measurement techniques, J. Geophys. Res.- http://www.atmos-chem-phys.net/8/5313/2008/. Atmos., 103, 22261–22280, 1998. Stull, R. B.: An Introduction to Boundary Layer Meteorology, Wood, E. C., Canagaratna, M., and Herndon, S.: Investigation of the Kluwer Academic Publishers, The Netherlands, 1988. correlation between odd-oxygen and secondary organic aerosol Thornton, J. A., Wooldridge, P. J., Cohen, R. C., Martinez, M., in Mexico City and Houston, in preparation, 2009. Harder, H., Brune, W. H., Williams, E. J., Roberts, J. M., Fehsen- Yauk, C., Polyzos, A., Rowan-Carroll, A., Somers, C. M., God- feld, F. C., Hall, S. R., Shetter, R. E., Wert, B. P., and Fried, A.: schalk, R. W., Van Schooten, F. J., Berndt, M. L., Pogribny, I. Ozone production rates as a function of NOx abundances and P., Koturbash, I., Williams, A., Douglas, G. R., and Kovalchuk, HOx production rates in the Nashville urban plume, J. Geophys. O.: Germ-line , DNA damage, and global hyperme- Res.-Atmos., 107, 4146, 2002. thylation in mice exposed to particulate air pollution in an ur- Thornton, J. A., Jaegle, L., and McNeill, V. F.: Assessing known ban/industrial location, Proceedings of the National Academy of pathways for HO2 loss in aqueous atmospheric aerosols: Re- Science, 105, 605–610, 2008. gional and global impacts on tropospheric oxidants, J. Geophys. Zavala, M., Herndon, S. C., Slott, R. S., Dunlea, E. J., Marr, L. C., Res., 113, 113, D05303, doi:10.1029/2007JD009236, 2008. Shorter, J. H., Zahniser, M., Knighton, W. B., Rogers, T. M., and Tie, X., Madronich, S., Li, G., Ying, Z., Zhang, R., Garcia, A. R., Kolb, C. E.: Characterization of on-road vehicle emissions in Lee-Taylor, J., and Liu, Y.: Characterizations of chemical ox- the Mexico City Metropolitan Area using a mobile laboratory in idants in Mexico City: A regional chemical dynamical model chase and fleet average measurement modes during the MCMA- (WRF-Chem) study, Atmos. Environ., 41, 1989–2008, 2007. 2003 field campaign, Atmos. Chem. Phys., 6, 5129–5142, 2006, Tressol, M., Ordonez, C., Zbinden, R., Brioude, J., Thouret, http://www.atmos-chem-phys.net/6/5129/2006/. V., Mari, C., Nedelec, P., Cammas, J. P., Smit, H., Patz, Zhao, J. and Zhang, R.: Proton transfer reaction rate constants be- H. W., and Volz-Thomas, A.: Air pollution during the 2003 tween hydronium ion (H3O+) and volatile organic compounds European heat wave as seen by MOZAIC airliners, Atmos. (VOCs), Atmos. Environ., 38, 2177–2185, 2004. Chem. Phys., 8, 2133–2150, 2008, http://www.atmos-chem- Zheng, J., Zhang, R., Fortner, E. C., Volkamer, R. M., Molina, phys.net/8/2133/2008/. L., Aiken, A. C., Jimenez, J. L., Gaggeler, K., Dommen, J., UN: World Prospects: The 2007 Revision, United Na- Dusanter, S., Stevens, P. S., and Tie, X.: Measurements of tions, New York, USA, 2008. HNO3 and N2O5 using Ion drift – Mass Velasco, E., Lamb, B., Westberg, H., Allwine, E., Sosa, G., Arriaga- Spectrometry during the MCMA – 2006 Campaign, Atmos. Colina, J. L., Jobson, B. T., Alexander, M. L., Prazeller, P., Chem. Phys., 8, 6823–6838, 2008, http://www.atmos-chem- Knighton, W. B., Rogers, T. M., Grutter, M., Herndon, S. C., phys.net/8/6823/2008/.

Atmos. Chem. Phys., 9, 2499–2517, 2009 www.atmos-chem-phys.net/9/2499/2009/